[DISCUSSION] Ignite integration testing framework.

[Max Shonichev]

Anton,

I've prepared a PoC of running Tiden in dockerized environment.

The code is in fork of your repo at
https://github.com/mshonichev/ignite.git, branch 'ignite-ducktape',
module 'integration-tests.


Steps to run PoC are as follows:
```
$ mkdir -p $HOME/tiden_poc
$ cd $HOME/tiden_poc
$ git clone -b ignite-ducktape https://github.com/mshonichev/ignite.git
$ cd ignite
$ scripts/build.sh
$ cd modules/integration-tests
$ mvn -DskipTests -Dmaven.javadoc.skip=true verify

```

Few changes in Tiden 0.6.5 include ability to set num_nodes upon
application start and minor fixes in API.

Most of the work to run in docker done by external bash scripts, so no
changes in tests themselves would be required to run them in bare-metal.

You can run them manually, or as a part of maven verify stage, they are
hooked in pom.xml via maven-exec-plugin for that.

I've noted comments in your PoC and instead of single docker image
we prepare a bunch of images:
  - tiden-master
  - tiden-slave:${JDK_VERSION}
  - tiden-artifacts-ignite:${IGNITE_VERSION}
  - tiden-artifacts-...

During provisioning stage, all those images are hotlinked via separate
volumes into /opt container dir. Tiden itself is installed as package
either of specific version or in 'develop' mode.

Also, it turns out that using `docker run --user=...` work great only in
MacOs or in Ubuntu under default user. Trying to run ducktests under
user with UID != 1000 produces unaccessible files in `results` dir. So,
in Tiden PoC I've fixed that too, you can review Dockerfile's in
modules/integration-tests/tiden/docker/ dir.

Next, I've recreated all your benchmarks, their code is in
modules/integration-tests/tiden/suites/benchmarks

Your Java applications copy-pasted with no changes into
modules/integration-tests/src, a thin Python wrapper over it is in
modules/integration-tests/apps/igniteaware.app

Unfortunately, no version-filtering decorators yet present, so instead
of single run across all versions, run-tests.sh internally runs Tiden
several times.


Please, checkout sources and share some thoughts.


On 09.07.2020 10:11, Max Shonichev wrote:

> Anton,
>
> well, strange thing, but clean up and rerun helped.
>
>
> Ubuntu 18.04
>
> ====================================================================================================
>
> SESSION REPORT (ALL TESTS)
> ducktape version: 0.7.7
> session_id:       2020-07-06--003
> run time:         4 minutes 44.835 seconds
> tests run:        5
> passed:           5
> failed:           0
> ignored:          0
> ====================================================================================================
>
> test_id:
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=2.8.1
>
> status:     PASS
> run time:   41.927 seconds
> {"Rebalanced in (sec)": 1.02205491065979}
> ----------------------------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=dev
>
> status:     PASS
> run time:   51.985 seconds
> {"Rebalanced in (sec)": 0.0760810375213623}
> ----------------------------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=2.7.6
>
> status:     PASS
> run time:   1 minute 4.283 seconds
> {"Streamed txs": "1900", "Measure duration (ms)": "34818", "Worst
> latency (ms)": "31035"}
> ----------------------------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=dev
>
> status:     PASS
> run time:   1 minute 13.089 seconds
> {"Streamed txs": "73134", "Measure duration (ms)": "35843", "Worst
> latency (ms)": "139"}
> ----------------------------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.spark_integration_test.SparkIntegrationTest.test_spark_client
>
> status:     PASS
> run time:   53.332 seconds
> ----------------------------------------------------------------------------------------------------
>
>
>
> MacBook
> ================================================================================
>
> SESSION REPORT (ALL TESTS)
> ducktape version: 0.7.7
> session_id:       2020-07-06--001
> run time:         6 minutes 58.612 seconds
> tests run:        5
> passed:           5
> failed:           0
> ignored:          0
> ================================================================================
>
> test_id:
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=2.8.1
>
> status:     PASS
> run time:   48.724 seconds
> {"Rebalanced in (sec)": 3.2574470043182373}
> --------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=dev
>
> status:     PASS
> run time:   1 minute 23.210 seconds
> {"Rebalanced in (sec)": 2.165921211242676}
> --------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=2.7.6
>
> status:     PASS
> run time:   1 minute 12.659 seconds
> {"Streamed txs": "642", "Measure duration (ms)": "33177", "Worst latency
> (ms)": "31063"}
> --------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=dev
>
> status:     PASS
> run time:   1 minute 57.257 seconds
> {"Streamed txs": "32924", "Measure duration (ms)": "48252", "Worst
> latency (ms)": "1010"}
> --------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.spark_integration_test.SparkIntegrationTest.test_spark_client
>
> status:     PASS
> run time:   1 minute 36.317 seconds
>
> =============
>
> while relative numbers proportion remains the same for different Ignite
> versions, absolute number for mac/linux differ more then twice.
>
> I'm finalizing code with 'local Tiden' appliance for your tests.  PR
> would be ready soon.
>
> Have you had a chance to deploy ducktests in bare metal?
>
>
>
> On 06.07.2020 14:27, Anton Vinogradov wrote:
>> Max,
>>
>> Thanks for the check!
>>
>>> Is it OK for those tests to fail?
>> No.
>> I see really strange things at logs.
>> Looks like you have concurrent ducktests run started not expected
>> services,
>> and this broke the tests.
>> Could you please clean up the docker (use clean-up script [1]).
>> Compile sources (use script [2]) and rerun the tests.
>>
>> [1]
>> https://github.com/anton-vinogradov/ignite/blob/dc98ee9df90b25eb5d928090b0e78b48cae2392e/modules/ducktests/tests/docker/clean_up.sh 
>>
>> [2]
>> https://github.com/anton-vinogradov/ignite/blob/3c39983005bd9eaf8cb458950d942fb592fff85c/scripts/build.sh 
>>
>>
>> On Mon, Jul 6, 2020 at 12:03 PM Nikolay Izhikov <[hidden email]>
>> wrote:
>>
>>> Hello, Maxim.
>>>
>>> Thanks for writing down the minutes.
>>>
>>> There is no such thing as «Nikolay team» on the dev-list.
>>> I propose to focus on product requirements and what we want to gain from
>>> the framework instead of taking into account the needs of some team.
>>>
>>> Can you, please, write down your version of requirements so we can
>>> reach a
>>> consensus on that and therefore move to the discussion of the
>>> implementation?
>>>
>>>> 6 июля 2020 г., в 11:18, Max Shonichev <[hidden email]> написал(а):
>>>>
>>>> Yes, Denis,
>>>>
>>>> common ground seems to be as follows:
>>>> Anton Vinogradov and Nikolay Izhikov would try to prepare and run PoC
>>> over physical hosts and share benchmark results. In the meantime,
>>> while I
>>> strongly believe that dockerized approach to benchmarking is a road to
>>> misleading and false positives, I'll prepare a PoC of Tiden in
>>> dockerized
>>> environment to support 'fast development prototyping' usecase Nikolay
>>> team
>>> insist on. It should be a matter of few days.
>>>>
>>>> As a side note, I've run Anton PoC locally and would like to have some
>>> comments about results:
>>>>
>>>> Test system: Ubuntu 18.04, docker 19.03.6
>>>> Test commands:
>>>>
>>>>
>>>> git clone -b ignite-ducktape [hidden email]:anton-vinogradov/ignite.git
>>>> cd ignite
>>>> mvn clean install -DskipTests -Dmaven.javadoc.skip=true
>>> -Pall-java,licenses,lgpl,examples,!spark-2.4,!spark,!scala
>>>> cd modules/ducktests/tests/docker
>>>> ./run_tests.sh
>>>>
>>>> Test results:
>>>>
>>> ====================================================================================================
>>>
>>>> SESSION REPORT (ALL TESTS)
>>>> ducktape version: 0.7.7
>>>> session_id:       2020-07-05--004
>>>> run time:         7 minutes 36.360 seconds
>>>> tests run:        5
>>>> passed:           3
>>>> failed:           2
>>>> ignored:          0
>>>>
>>> ====================================================================================================
>>>
>>>> test_id:
>>> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=2.8.1
>>>
>>>> status:     FAIL
>>>> run time:   3 minutes 12.232 seconds
>>>>
>>> ----------------------------------------------------------------------------------------------------
>>>
>>>> test_id:
>>> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=2.7.6
>>>
>>>> status:     FAIL
>>>> run time:   1 minute 33.076 seconds
>>>>
>>>>
>>>> Is it OK for those tests to fail? Attached is full test report
>>>>
>>>>
>>>> On 02.07.2020 17:46, Denis Magda wrote:
>>>>> Folks,
>>>>> Please share the summary of that Slack conversation here for records
>>> once
>>>>> you find common ground.
>>>>> -
>>>>> Denis
>>>>> On Thu, Jul 2, 2020 at 3:22 AM Nikolay Izhikov <[hidden email]>
>>> wrote:
>>>>>> Igniters.
>>>>>>
>>>>>> All who are interested in integration testing framework discussion
>>>>>> are
>>>>>> welcome into slack channel -
>>>>>>
>>> https://join.slack.com/share/zt-fk2ovehf-TcomEAwiXaPzLyNKZbmfzw?cdn_fallback=2 
>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>> 2 июля 2020 г., в 13:06, Anton Vinogradov <[hidden email]>
>>>>>>> написал(а):
>>>>>>>
>>>>>>> Max,
>>>>>>> Thanks for joining us.
>>>>>>>
>>>>>>>> 1. tiden can deploy artifacts by itself, while ducktape relies on
>>>>>>>> dependencies being deployed by external scripts.
>>>>>>> No. It is important to distinguish development, deploy, and
>>>>>> orchestration.
>>>>>>> All-in-one solutions have extremely limited usability.
>>>>>>> As to Ducktests:
>>>>>>> Docker is responsible for deployments during development.
>>>>>>> CI/CD is responsible for deployments during release and nightly
>>> checks.
>>>>>> It's up to the team to chose AWS, VM, BareMetal, and even OS.
>>>>>>> Ducktape is responsible for orchestration.
>>>>>>>
>>>>>>>> 2. tiden can execute actions over remote nodes in real parallel
>>>>>> fashion,
>>>>>>>> while ducktape internally does all actions sequentially.
>>>>>>> No. Ducktape may start any service in parallel. See Pme-free
>>>>>>> benchmark
>>>>>> [1] for details.
>>>>>>>
>>>>>>>> if we used ducktape solution we would have to instead prepare some
>>>>>>>> deployment scripts to pre-initialize Sberbank hosts, for example,
>>> with
>>>>>>>> Ansible or Chef.
>>>>>>> Sure, because a way of deploy depends on infrastructure.
>>>>>>> How can we be sure that OS we use and the restrictions we have
>>>>>>> will be
>>>>>> compatible with Tiden?
>>>>>>>
>>>>>>>> You have solved this deficiency with docker by putting all
>>> dependencies
>>>>>>>> into one uber-image ...
>>>>>>> and
>>>>>>>> I guess we all know about docker hyped ability to run over
>>> distributed
>>>>>>>> virtual networks.
>>>>>>> It is very important not to confuse the test's development (docker
>>> image
>>>>>> you're talking about) and real deployment.
>>>>>>>
>>>>>>>> If we had stopped and started 5 nodes one-by-one, as ducktape does
>>>>>>> All actions can be performed in parallel.
>>>>>>> See how Ducktests [2] starts cluster in parallel for example.
>>>>>>>
>>>>>>> [1]
>>>>>>
>>> https://github.com/apache/ignite/pull/7967/files#diff-59adde2a2ab7dc17aea6c65153dfcda7R84 
>>>
>>>>>>> [2]
>>>>>>
>>> https://github.com/apache/ignite/pull/7967/files#diff-d6a7b19f30f349d426b8894a40389cf5R79 
>>>
>>>>>>>
>>>>>>> On Thu, Jul 2, 2020 at 1:00 PM Nikolay Izhikov <[hidden email]>
>>>>>> wrote:
>>>>>>> Hello, Maxim.
>>>>>>>
>>>>>>>> 1. tiden can deploy artifacts by itself, while ducktape relies on
>>>>>> dependencies being deployed by external scripts
>>>>>>>
>>>>>>> Why do you think that maintaining deploy scripts coupled with the
>>>>>> testing framework is an advantage?
>>>>>>> I thought we want to see and maintain deployment scripts separate
>>>>>>> from
>>>>>> the testing framework.
>>>>>>>
>>>>>>>> 2. tiden can execute actions over remote nodes in real parallel
>>>>>> fashion, while ducktape internally does all actions sequentially.
>>>>>>>
>>>>>>> Can you, please, clarify, what actions do you have in mind?
>>>>>>> And why we want to execute them concurrently?
>>>>>>> Ignite node start, Client application execution can be done
>>> concurrently
>>>>>> with the ducktape approach.
>>>>>>>
>>>>>>>> If we used ducktape solution we would have to instead prepare some
>>>>>> deployment scripts to pre-initialize Sberbank hosts, for example,
>>>>>> with
>>>>>> Ansible or Chef
>>>>>>>
>>>>>>> We shouldn’t take some user approach as an argument in this
>>> discussion.
>>>>>> Let’s discuss a general approach for all users of the Ignite. Anyway,
>>> what
>>>>>> is wrong with the external deployment script approach?
>>>>>>>
>>>>>>> We, as a community, should provide several ways to run integration
>>> tests
>>>>>> out-of-the-box AND the ability to customize deployment regarding the
>>> user
>>>>>> landscape.
>>>>>>>
>>>>>>>> You have solved this deficiency with docker by putting all
>>>>>> dependencies into one uber-image and that looks like simple and
>>>>>> elegant
>>>>>> solution however, that effectively limits you to single-host testing.
>>>>>>>
>>>>>>> Docker image should be used only by the Ignite developers to test
>>>>>> something locally.
>>>>>>> It’s not intended for some real-world testing.
>>>>>>>
>>>>>>> The main issue with the Tiden that I see, it tested and
>>>>>>> maintained as
>>> a
>>>>>> closed source solution.
>>>>>>> This can lead to the hard to solve problems when we start using and
>>>>>> maintaining it as an open-source solution.
>>>>>>> Like, how many developers used Tiden? And how many of developers
>>>>>>> were
>>>>>> not authors of the Tiden itself?
>>>>>>>
>>>>>>>
>>>>>>>> 2 июля 2020 г., в 12:30, Max Shonichev <[hidden email]>
>>>>>> написал(а):
>>>>>>>>
>>>>>>>> Anton, Nikolay,
>>>>>>>>
>>>>>>>> Let's agree on what we are arguing about: whether it is about "like
>>> or
>>>>>> don't like" or about technical properties of suggested solutions.
>>>>>>>>
>>>>>>>> If it is about likes and dislikes, then the whole discussion is
>>>>>> meaningless. However, I hope together we can analyse pros and cons
>>>>>> carefully.
>>>>>>>>
>>>>>>>> As far as I can understand now, two main differences between
>>>>>>>> ducktape
>>>>>> and tiden is that:
>>>>>>>>
>>>>>>>> 1. tiden can deploy artifacts by itself, while ducktape relies on
>>>>>> dependencies being deployed by external scripts.
>>>>>>>>
>>>>>>>> 2. tiden can execute actions over remote nodes in real parallel
>>>>>> fashion, while ducktape internally does all actions sequentially.
>>>>>>>>
>>>>>>>> As for me, these are very important properties for distributed
>>> testing
>>>>>> framework.
>>>>>>>>
>>>>>>>> First property let us easily reuse tiden in existing
>>>>>>>> infrastructures,
>>>>>> for example, during Zookeeper IEP testing at Sberbank site we used
>>>>>> the
>>> same
>>>>>> tiden scripts that we use in our lab, the only change was putting a
>>> list of
>>>>>> hosts into config.
>>>>>>>>
>>>>>>>> If we used ducktape solution we would have to instead prepare some
>>>>>> deployment scripts to pre-initialize Sberbank hosts, for example,
>>>>>> with
>>>>>> Ansible or Chef.
>>>>>>>>
>>>>>>>>
>>>>>>>> You have solved this deficiency with docker by putting all
>>>>>> dependencies into one uber-image and that looks like simple and
>>>>>> elegant
>>>>>> solution,
>>>>>>>> however, that effectively limits you to single-host testing.
>>>>>>>>
>>>>>>>> I guess we all know about docker hyped ability to run over
>>> distributed
>>>>>> virtual networks. We used to go that way, but quickly found that
>>>>>> it is
>>> more
>>>>>> of the hype than real work. In real environments, there are problems
>>> with
>>>>>> routing, DNS, multicast and broadcast traffic, and many others, that
>>> turn
>>>>>> docker-based distributed solution into a fragile hard-to-maintain
>>> monster.
>>>>>>>>
>>>>>>>> Please, if you believe otherwise, perform a run of your PoC over at
>>>>>> least two physical hosts and share results with us.
>>>>>>>>
>>>>>>>> If you consider that one physical docker host is enough, please,
>>> don't
>>>>>> overlook that we want to run real scale scenarios, with 50-100 cache
>>>>>> groups, persistence enabled and a millions of keys loaded.
>>>>>>>>
>>>>>>>> Practical limit for such configurations is 4-6 nodes per single
>>>>>> physical host. Otherwise, tests become flaky due to resource
>>> starvation.
>>>>>>>>
>>>>>>>> Please, if you believe otherwise, perform at least a 10 of runs of
>>>>>> your PoC with other tests running at TC (we're targeting TeamCity,
>>> right?)
>>>>>> and share results so we could check if the numbers are reproducible.
>>>>>>>>
>>>>>>>> I stress this once more: functional integration tests are OK to run
>>> in
>>>>>> Docker and CI, but running benchmarks in Docker is a big NO GO.
>>>>>>>>
>>>>>>>>
>>>>>>>> Second property let us write tests that require real-parallel
>>>>>>>> actions
>>>>>> over hosts.
>>>>>>>>
>>>>>>>> For example, agreed scenario for PME benchmarkduring "PME
>>> optimization
>>>>>> stream" was as follows:
>>>>>>>>
>>>>>>>>   - 10 server nodes, preloaded with 1M of keys
>>>>>>>>   - 4 client nodes perform transactional load  (client nodes
>>> physically
>>>>>> separated from server nodes)
>>>>>>>>   - during load:
>>>>>>>>   -- 5 server nodes stopped in parallel
>>>>>>>>   -- after 1 minute, all 5 nodes are started in parallel
>>>>>>>>   - load stopped, logs are analysed for exchange times.
>>>>>>>>
>>>>>>>> If we had stopped and started 5 nodes one-by-one, as ducktape does,
>>>>>> then partition map exchange merge would not happen and we could not
>>> have
>>>>>> measured PME optimizations for that case.
>>>>>>>>
>>>>>>>>
>>>>>>>> These are limitations of ducktape that we believe as a more
>>>>>>>> important
>>>>>>>> argument "against" than you provide "for".
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> On 30.06.2020 14:58, Anton Vinogradov wrote:
>>>>>>>>> Folks,
>>>>>>>>> First, I've created PR [1] with ducktests improvements
>>>>>>>>> PR contains the following changes
>>>>>>>>> - Pme-free switch proof-benchmark (2.7.6 vs master)
>>>>>>>>> - Ability to check (compare with) previous releases (eg. 2.7.6 &
>>> 2.8)
>>>>>>>>> - Global refactoring
>>>>>>>>> -- benchmarks javacode simplification
>>>>>>>>> -- services python and java classes code deduplication
>>>>>>>>> -- fail-fast checks for java and python (eg. application should
>>>>>> explicitly write it finished with success)
>>>>>>>>> -- simple results extraction from tests and benchmarks
>>>>>>>>> -- javacode now configurable from tests/benchmarks
>>>>>>>>> -- proper SIGTERM handling at javacode (eg. it may finish last
>>>>>> operation and log results)
>>>>>>>>> -- docker volume now marked as delegated to increase execution
>>>>>>>>> speed
>>>>>> for mac & win users
>>>>>>>>> -- Ignite cluster now start in parallel (start speed-up)
>>>>>>>>> -- Ignite can be configured at test/benchmark
>>>>>>>>> - full and module assembly scripts added
>>>>>>>> Great job done! But let me remind one of Apache Ignite principles:
>>>>>>>> week of thinking save months of development.
>>>>>>>>
>>>>>>>>
>>>>>>>>> Second, I'd like to propose to accept ducktests [2] (ducktape
>>>>>> integration) as a target "PoC check & real topology benchmarking
>>>>>> tool".
>>>>>>>>> Ducktape pros
>>>>>>>>> - Developed for distributed system by distributed system
>>>>>>>>> developers.
>>>>>>>> So does Tiden
>>>>>>>>
>>>>>>>>> - Developed since 2014, stable.
>>>>>>>> Tiden is also pretty stable, and development start date is not a
>>>>>>>> good
>>>>>> argument, for example pytest is since 2004, pytest-xdist (plugin for
>>>>>> distributed testing) is since 2010, but we don't see it as a
>>> alternative at
>>>>>> all.
>>>>>>>>
>>>>>>>>> - Proven usability by usage at Kafka.
>>>>>>>> Tiden is proven usable by usage at GridGain and Sberbank
>>>>>>>> deployments.
>>>>>>>> Core, storage, sql and tx teams use benchmark results provided by
>>>>>> Tiden on a daily basis.
>>>>>>>>
>>>>>>>>> - Dozens of dozens tests and benchmarks at Kafka as a great
>>>>>>>>> example
>>>>>> pack.
>>>>>>>> We'll donate some of our suites to Ignite as I've mentioned in
>>>>>> previous letter.
>>>>>>>>
>>>>>>>>> - Built-in Docker support for rapid development and checks.
>>>>>>>> False, there's no specific 'docker support' in ducktape itself, you
>>>>>> just wrap it in docker by yourself, because ducktape is lacking
>>> deployment
>>>>>> abilities.
>>>>>>>>
>>>>>>>>> - Great for CI automation.
>>>>>>>> False, there's no specific CI-enabled features in ducktape.
>>>>>>>> Tiden, on
>>>>>> the other hand, provide generic xUnit reporting format, which is
>>> supported
>>>>>> by both TeamCity and Jenkins. Also, instead of using private keys,
>>> Tiden
>>>>>> can use SSH agent, which is also great for CI, because both
>>>>>>>> TeamCity and Jenkins store keys in secret storage available only
>>>>>>>> for
>>>>>> ssh-agent and only for the time of the test.
>>>>>>>>
>>>>>>>>
>>>>>>>>>> As an additional motivation, at least 3 teams
>>>>>>>>> - IEP-45 team (to check crash-recovery speed-up (discovery and
>>> Zabbix
>>>>>> speed-up))
>>>>>>>>> - Ignite SE Plugins team (to check plugin's features does not
>>>>>> slow-down or broke AI features)
>>>>>>>>> - Ignite SE QA team (to append already developed
>>>>>>>>> smoke/load/failover
>>>>>> tests to AI codebase)
>>>>>>>>
>>>>>>>> Please, before recommending your tests to other teams, provide
>>>>>>>> proofs
>>>>>>>> that your tests are reproducible in real environment.
>>>>>>>>
>>>>>>>>
>>>>>>>>> now, wait for ducktest merge to start checking cases they
>>>>>>>>> working on
>>>>>> in AI way.
>>>>>>>>> Thoughts?
>>>>>>>> Let us together review both solutions, we'll try to run your
>>>>>>>> tests in
>>>>>> our lab, and you'll try to at least checkout tiden and see if same
>>> tests
>>>>>> can be implemented with it?
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>> [1] https://github.com/apache/ignite/pull/7967
>>>>>>>>> [2] https://github.com/apache/ignite/tree/ignite-ducktape
>>>>>>>>> On Tue, Jun 16, 2020 at 12:22 PM Nikolay Izhikov <
>>> [hidden email]
>>>>>> <mailto:[hidden email]>> wrote:
>>>>>>>>>     Hello, Maxim.
>>>>>>>>>     Thank you for so detailed explanation.
>>>>>>>>>     Can we put the content of this discussion somewhere on the
>>>>>>>>> wiki?
>>>>>>>>>     So It doesn’t get lost.
>>>>>>>>>     I divide the answer in several parts. From the requirements to
>>> the
>>>>>>>>>     implementation.
>>>>>>>>>     So, if we agreed on the requirements we can proceed with the
>>>>>>>>>     discussion of the implementation.
>>>>>>>>>     1. Requirements:
>>>>>>>>>     The main goal I want to achieve is *reproducibility* of the
>>> tests.
>>>>>>>>>     I’m sick and tired with the zillions of flaky, rarely
>>>>>>>>> failed, and
>>>>>>>>>     almost never failed tests in Ignite codebase.
>>>>>>>>>     We should start with the simplest scenarios that will be as
>>>>>> reliable
>>>>>>>>>     as steel :)
>>>>>>>>>     I want to know for sure:
>>>>>>>>>        - Is this PR makes rebalance quicker or not?
>>>>>>>>>        - Is this PR makes PME quicker or not?
>>>>>>>>>     So, your description of the complex test scenario looks as
>>>>>>>>> a next
>>>>>>>>>     step to me.
>>>>>>>>>     Anyway, It’s cool we already have one.
>>>>>>>>>     The second goal is to have a strict test lifecycle as we
>>>>>>>>> have in
>>>>>>>>>     JUnit and similar frameworks.
>>>>>>>>>      > It covers production-like deployment and running a
>>>>>>>>> scenarios
>>>>>> over
>>>>>>>>>     a single database instance.
>>>>>>>>>     Do you mean «single cluster» or «single host»?
>>>>>>>>>     2. Existing tests:
>>>>>>>>>      > A Combinator suite allows to run set of operations
>>> concurrently
>>>>>>>>>     over given database instance.
>>>>>>>>>      > A Consumption suite allows to run a set production-like
>>> actions
>>>>>>>>>     over given set of Ignite/GridGain versions and compare test
>>> metrics
>>>>>>>>>     across versions
>>>>>>>>>      > A Yardstick suite
>>>>>>>>>      > A Stress suite that simulates hardware environment
>>>>>>>>> degradation
>>>>>>>>>      > An Ultimate, DR and Compatibility suites that performs
>>>>>> functional
>>>>>>>>>     regression testing
>>>>>>>>>      > Regression
>>>>>>>>>     Great news that we already have so many choices for testing!
>>>>>>>>>     Mature test base is a big +1 for Tiden.
>>>>>>>>>     3. Comparison:
>>>>>>>>>      > Criteria: Test configuration
>>>>>>>>>      > Ducktape: single JSON string for all tests
>>>>>>>>>      > Tiden: any number of YaML config files, command line option
>>> for
>>>>>>>>>     fine-grained test configuration, ability to select/modify
>>>>>>>>> tests
>>>>>>>>>     behavior based on Ignite version.
>>>>>>>>>     1. Many YAML files can be hard to maintain.
>>>>>>>>>     2. In ducktape, you can set parameters via «—parameters»
>>>>>>>>> option.
>>>>>>>>>     Please, take a look at the doc [1]
>>>>>>>>>      > Criteria: Cluster control
>>>>>>>>>      > Tiden: additionally can address cluster as a whole and
>>>>>>>>> execute
>>>>>>>>>     remote commands in parallel.
>>>>>>>>>     It seems we implement this ability in the PoC, already.
>>>>>>>>>      > Criteria: Test assertions
>>>>>>>>>      > Tiden: simple asserts, also few customized assertion
>>>>>>>>> helpers.
>>>>>>>>>      > Ducktape: simple asserts.
>>>>>>>>>     Can you, please, be more specific.
>>>>>>>>>     What helpers do you have in mind?
>>>>>>>>>     Ducktape has an asserts that waits for logfile messages or
>>>>>>>>> some
>>>>>>>>>     process finish.
>>>>>>>>>      > Criteria: Test reporting
>>>>>>>>>      > Ducktape: limited to its own text/HTML format
>>>>>>>>>     Ducktape have
>>>>>>>>>     1. Text reporter
>>>>>>>>>     2. Customizable HTML reporter
>>>>>>>>>     3. JSON reporter.
>>>>>>>>>     We can show JSON with the any template or tool.
>>>>>>>>>      > Criteria: Provisioning and deployment
>>>>>>>>>      > Ducktape: can provision subset of hosts from cluster for
>>>>>>>>> test
>>>>>>>>>     needs. However, that means, that test can’t be scaled without
>>> test
>>>>>>>>>     code changes. Does not do any deploy, relies on external
>>>>>>>>> means,
>>>>>> e.g.
>>>>>>>>>     pre-packaged in docker image, as in PoC.
>>>>>>>>>     This is not true.
>>>>>>>>>     1. We can set explicit test parameters(node number) via
>>> parameters.
>>>>>>>>>     We can increase client count of cluster size without test code
>>>>>> changes.
>>>>>>>>>     2. We have many choices for the test environment. These
>>>>>>>>> choices
>>> are
>>>>>>>>>     tested and used in other projects:
>>>>>>>>>              * docker
>>>>>>>>>              * vagrant
>>>>>>>>>              * private cloud(ssh access)
>>>>>>>>>              * ec2
>>>>>>>>>     Please, take a look at Kafka documentation [2]
>>>>>>>>>      > I can continue more on this, but it should be enough for
>>>>>>>>> now:
>>>>>>>>>     We need to go deeper! :)
>>>>>>>>>     [1]
>>>>>>>>>
>>>>>> https://ducktape-docs.readthedocs.io/en/latest/run_tests.html#options
>>>>>>>>>     [2]
>>>>>> https://github.com/apache/kafka/tree/trunk/tests#ec2-quickstart
>>>>>>>>>      > 9 июня 2020 г., в 17:25, Max A. Shonichev
>>>>>>>>> <[hidden email]
>>>>>>>>>     <mailto:[hidden email]>> написал(а):
>>>>>>>>>      >
>>>>>>>>>      > Greetings, Nikolay,
>>>>>>>>>      >
>>>>>>>>>      > First of all, thank you for you great effort preparing
>>>>>>>>> PoC of
>>>>>>>>>     integration testing to Ignite community.
>>>>>>>>>      >
>>>>>>>>>      > It’s a shame Ignite did not have at least some such
>>>>>>>>> tests yet,
>>>>>>>>>     however, GridGain, as a major contributor to Apache Ignite
>>>>>>>>> had a
>>>>>>>>>     profound collection of in-house tools to perform
>>>>>>>>> integration and
>>>>>>>>>     performance testing for years already and while we slowly
>>> consider
>>>>>>>>>     sharing our expertise with the community, your initiative
>>>>>>>>> makes
>>> us
>>>>>>>>>     drive that process a bit faster, thanks a lot!
>>>>>>>>>      >
>>>>>>>>>      > I reviewed your PoC and want to share a little about
>>>>>>>>> what we
>>> do
>>>>>>>>>     on our part, why and how, hope it would help community take
>>> proper
>>>>>>>>>     course.
>>>>>>>>>      >
>>>>>>>>>      > First I’ll do a brief overview of what decisions we made
>>>>>>>>> and
>>>>>> what
>>>>>>>>>     we do have in our private code base, next I’ll describe
>>>>>>>>> what we
>>>>>> have
>>>>>>>>>     already donated to the public and what we plan public next,
>>>>>>>>> then
>>>>>>>>>     I’ll compare both approaches highlighting deficiencies in
>>>>>>>>> order
>>> to
>>>>>>>>>     spur public discussion on the matter.
>>>>>>>>>      >
>>>>>>>>>      > It might seem strange to use Python to run Bash to run Java
>>>>>>>>>     applications because that introduces IT industry best of
>>>>>>>>> breed’ –
>>>>>>>>>     the Python dependency hell – to the Java application code
>>>>>>>>> base.
>>> The
>>>>>>>>>     only strangest decision one can made is to use Maven to run
>>> Docker
>>>>>>>>>     to run Bash to run Python to run Bash to run Java, but
>>>>>>>>> desperate
>>>>>>>>>     times call for desperate measures I guess.
>>>>>>>>>      >
>>>>>>>>>      > There are Java-based solutions for integration testing
>>>>>>>>> exists,
>>>>>>>>>     e.g. Testcontainers [1], Arquillian [2], etc, and they
>>>>>>>>> might go
>>>>>> well
>>>>>>>>>     for Ignite community CI pipelines by them selves. But we also
>>>>>> wanted
>>>>>>>>>     to run performance tests and benchmarks, like the dreaded PME
>>>>>>>>>     benchmark, and this is solved by totally different set of
>>>>>>>>> tools
>>> in
>>>>>>>>>     Java world, e.g. Jmeter [3], OpenJMH [4], Gatling [5], etc.
>>>>>>>>>      >
>>>>>>>>>      > Speaking specifically about benchmarking, Apache Ignite
>>>>>> community
>>>>>>>>>     already has Yardstick [6], and there’s nothing wrong with
>>>>>>>>> writing
>>>>>>>>>     PME benchmark using Yardstick, but we also wanted to be
>>>>>>>>> able to
>>> run
>>>>>>>>>     scenarios like this:
>>>>>>>>>      > - put an X load to a Ignite database;
>>>>>>>>>      > - perform an Y set of operations to check how Ignite copes
>>> with
>>>>>>>>>     operations under load.
>>>>>>>>>      >
>>>>>>>>>      > And yes, we also wanted applications under test be deployed
>>>>>> ‘like
>>>>>>>>>     in a production’, e.g. distributed over a set of hosts. This
>>> arises
>>>>>>>>>     questions about provisioning and nodes affinity which I’ll
>>>>>>>>> cover
>>> in
>>>>>>>>>     detail later.
>>>>>>>>>      >
>>>>>>>>>      > So we decided to put a little effort to build a simple
>>>>>>>>> tool to
>>>>>>>>>     cover different integration and performance scenarios, and
>>>>>>>>> our QA
>>>>>>>>>     lab first attempt was PoC-Tester [7], currently open source
>>>>>>>>> for
>>> all
>>>>>>>>>     but for reporting web UI. It’s a quite simple to use 95%
>>> Java-based
>>>>>>>>>     tool targeted to be run on a pre-release QA stage.
>>>>>>>>>      >
>>>>>>>>>      > It covers production-like deployment and running a
>>>>>>>>> scenarios
>>>>>> over
>>>>>>>>>     a single database instance. PoC-Tester scenarios consists of a
>>>>>>>>>     sequence of tasks running sequentially or in parallel.
>>>>>>>>> After all
>>>>>>>>>     tasks complete, or at any time during test, user can run logs
>>>>>>>>>     collection task, logs are checked against exceptions and a
>>> summary
>>>>>>>>>     of found issues and task ops/latency statistics is
>>>>>>>>> generated at
>>> the
>>>>>>>>>     end of scenario. One of the main PoC-Tester features is its
>>>>>>>>>     fire-and-forget approach to task managing. That is, you can
>>> deploy
>>>>>> a
>>>>>>>>>     grid and left it running for weeks, periodically firing some
>>> tasks
>>>>>>>>>     onto it.
>>>>>>>>>      >
>>>>>>>>>      > During earliest stages of PoC-Tester development it becomes
>>>>>> quite
>>>>>>>>>     clear that Java application development is a tedious
>>>>>>>>> process and
>>>>>>>>>     architecture decisions you take during development are slow
>>>>>>>>> and
>>>>>> hard
>>>>>>>>>     to change.
>>>>>>>>>      > For example, scenarios like this
>>>>>>>>>      > - deploy two instances of GridGain with master-slave data
>>>>>>>>>     replication configured;
>>>>>>>>>      > - put a load on master;
>>>>>>>>>      > - perform checks on slave,
>>>>>>>>>      > or like this:
>>>>>>>>>      > - preload a 1Tb of data by using your favorite tool of
>>>>>>>>> choice
>>> to
>>>>>>>>>     an Apache Ignite of version X;
>>>>>>>>>      > - run a set of functional tests running Apache Ignite
>>>>>>>>> version
>>> Y
>>>>>>>>>     over preloaded data,
>>>>>>>>>      > do not fit well in the PoC-Tester workflow.
>>>>>>>>>      >
>>>>>>>>>      > So, this is why we decided to use Python as a generic
>>> scripting
>>>>>>>>>     language of choice.
>>>>>>>>>      >
>>>>>>>>>      > Pros:
>>>>>>>>>      > - quicker prototyping and development cycles
>>>>>>>>>      > - easier to find DevOps/QA engineer with Python skills than
>>> one
>>>>>>>>>     with Java skills
>>>>>>>>>      > - used extensively all over the world for DevOps/CI
>>>>>>>>> pipelines
>>>>>> and
>>>>>>>>>     thus has rich set of libraries for all possible integration
>>>>>>>>> uses
>>>>>> cases.
>>>>>>>>>      >
>>>>>>>>>      > Cons:
>>>>>>>>>      > - Nightmare with dependencies. Better stick to specific
>>>>>>>>>     language/libraries version.
>>>>>>>>>      >
>>>>>>>>>      > Comparing alternatives for Python-based testing
>>>>>>>>> framework we
>>>>>> have
>>>>>>>>>     considered following requirements, somewhat similar to what
>>> you’ve
>>>>>>>>>     mentioned for Confluent [8] previously:
>>>>>>>>>      > - should be able run locally or distributed (bare metal
>>>>>>>>> or in
>>>>>> the
>>>>>>>>>     cloud)
>>>>>>>>>      > - should have built-in deployment facilities for
>>>>>>>>> applications
>>>>>>>>>     under test
>>>>>>>>>      > - should separate test configuration and test code
>>>>>>>>>      > -- be able to easily reconfigure tests by simple
>>>>>>>>> configuration
>>>>>>>>>     changes
>>>>>>>>>      > -- be able to easily scale test environment by simple
>>>>>>>>>     configuration changes
>>>>>>>>>      > -- be able to perform regression testing by simple
>>>>>>>>> switching
>>>>>>>>>     artifacts under test via configuration
>>>>>>>>>      > -- be able to run tests with different JDK version by
>>>>>>>>> simple
>>>>>>>>>     configuration changes
>>>>>>>>>      > - should have human readable reports and/or reporting tools
>>>>>>>>>     integration
>>>>>>>>>      > - should allow simple test progress monitoring, one does
>>>>>>>>> not
>>>>>> want
>>>>>>>>>     to run 6-hours test to find out that application actually
>>>>>>>>> crashed
>>>>>>>>>     during first hour.
>>>>>>>>>      > - should allow parallel execution of test actions
>>>>>>>>>      > - should have clean API for test writers
>>>>>>>>>      > -- clean API for distributed remote commands execution
>>>>>>>>>      > -- clean API for deployed applications start / stop and
>>>>>>>>> other
>>>>>>>>>     operations
>>>>>>>>>      > -- clean API for performing check on results
>>>>>>>>>      > - should be open source or at least source code should
>>>>>>>>> allow
>>>>>> ease
>>>>>>>>>     change or extension
>>>>>>>>>      >
>>>>>>>>>      > Back at that time we found no better alternative than to
>>>>>>>>> write
>>>>>>>>>     our own framework, and here goes Tiden [9] as GridGain
>>>>>>>>> framework
>>> of
>>>>>>>>>     choice for functional integration and performance testing.
>>>>>>>>>      >
>>>>>>>>>      > Pros:
>>>>>>>>>      > - solves all the requirements above
>>>>>>>>>      > Cons (for Ignite):
>>>>>>>>>      > - (currently) closed GridGain source
>>>>>>>>>      >
>>>>>>>>>      > On top of Tiden we’ve built a set of test suites, some of
>>> which
>>>>>>>>>     you might have heard already.
>>>>>>>>>      >
>>>>>>>>>      > A Combinator suite allows to run set of operations
>>> concurrently
>>>>>>>>>     over given database instance. Proven to find at least 30+ race
>>>>>>>>>     conditions and NPE issues.
>>>>>>>>>      >
>>>>>>>>>      > A Consumption suite allows to run a set production-like
>>> actions
>>>>>>>>>     over given set of Ignite/GridGain versions and compare test
>>> metrics
>>>>>>>>>     across versions, like heap/disk/CPU consumption, time to
>>>>>>>>> perform
>>>>>>>>>     actions, like client PME, server PME, rebalancing time, data
>>>>>>>>>     replication time, etc.
>>>>>>>>>      >
>>>>>>>>>      > A Yardstick suite is a thin layer of Python glue code to
>>>>>>>>> run
>>>>>>>>>     Apache Ignite pre-release benchmarks set. Yardstick itself
>>>>>>>>> has a
>>>>>>>>>     mediocre deployment capabilities, Tiden solves this easily.
>>>>>>>>>      >
>>>>>>>>>      > A Stress suite that simulates hardware environment
>>>>>>>>> degradation
>>>>>>>>>     during testing.
>>>>>>>>>      >
>>>>>>>>>      > An Ultimate, DR and Compatibility suites that performs
>>>>>> functional
>>>>>>>>>     regression testing of GridGain Ultimate Edition features like
>>>>>>>>>     snapshots, security, data replication, rolling upgrades, etc.
>>>>>>>>>      >
>>>>>>>>>      > A Regression and some IEPs testing suites, like IEP-14,
>>> IEP-15,
>>>>>>>>>     etc, etc, etc.
>>>>>>>>>      >
>>>>>>>>>      > Most of the suites above use another in-house developed
>>>>>>>>> Java
>>>>>> tool
>>>>>>>>>     – PiClient – to perform actual loading and miscellaneous
>>> operations
>>>>>>>>>     with Ignite under test. We use py4j Python-Java gateway
>>>>>>>>> library
>>> to
>>>>>>>>>     control PiClient instances from the tests.
>>>>>>>>>      >
>>>>>>>>>      > When we considered CI, we put TeamCity out of scope,
>>>>>>>>> because
>>>>>>>>>     distributed integration and performance tests tend to run for
>>> hours
>>>>>>>>>     and TeamCity agents are scarce and costly resource. So,
>>>>>>>>> bundled
>>>>>> with
>>>>>>>>>     Tiden there is jenkins-job-builder [10] based CI pipelines and
>>>>>>>>>     Jenkins xUnit reporting. Also, rich web UI tool Ward
>>>>>>>>> aggregates
>>>>>> test
>>>>>>>>>     run reports across versions and has built in visualization
>>> support
>>>>>>>>>     for Combinator suite.
>>>>>>>>>      >
>>>>>>>>>      > All of the above is currently closed source, but we plan to
>>> make
>>>>>>>>>     it public for community, and publishing Tiden core [9] is the
>>> first
>>>>>>>>>     step on that way. You can review some examples of using
>>>>>>>>> Tiden for
>>>>>>>>>     tests at my repository [11], for start.
>>>>>>>>>      >
>>>>>>>>>      > Now, let’s compare Ducktape PoC and Tiden.
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Language
>>>>>>>>>      > Tiden: Python, 3.7
>>>>>>>>>      > Ducktape: Python, proposes itself as Python 2.7, 3.6, 3.7
>>>>>>>>>     compatible, but actually can’t work with Python 3.7 due to
>>>>>>>>> broken
>>>>>>>>>     Zmq dependency.
>>>>>>>>>      > Comment: Python 3.7 has a much better support for
>>>>>>>>> async-style
>>>>>>>>>     code which might be crucial for distributed application
>>>>>>>>> testing.
>>>>>>>>>      > Score: Tiden: 1, Ducktape: 0
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Test writers API
>>>>>>>>>      > Supported integration test framework concepts are basically
>>> the
>>>>>> same:
>>>>>>>>>      > - a test controller (test runner)
>>>>>>>>>      > - a cluster
>>>>>>>>>      > - a node
>>>>>>>>>      > - an application (a service in Ducktape terms)
>>>>>>>>>      > - a test
>>>>>>>>>      > Score: Tiden: 5, Ducktape: 5
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Tests selection and run
>>>>>>>>>      > Ducktape: suite-package-class-method level selection,
>>>>>>>>> internal
>>>>>>>>>     scheduler allows to run tests in suite in parallel.
>>>>>>>>>      > Tiden: also suite-package-class-method level selection,
>>>>>>>>>     additionally allows selecting subset of tests by attribute,
>>>>>> parallel
>>>>>>>>>     runs not built in, but allows merging test reports after
>>> different
>>>>>> runs.
>>>>>>>>>      > Score: Tiden: 2, Ducktape: 2
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Test configuration
>>>>>>>>>      > Ducktape: single JSON string for all tests
>>>>>>>>>      > Tiden: any number of YaML config files, command line option
>>> for
>>>>>>>>>     fine-grained test configuration, ability to select/modify
>>>>>>>>> tests
>>>>>>>>>     behavior based on Ignite version.
>>>>>>>>>      > Score: Tiden: 3, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Cluster control
>>>>>>>>>      > Ducktape: allow execute remote commands by node granularity
>>>>>>>>>      > Tiden: additionally can address cluster as a whole and
>>>>>>>>> execute
>>>>>>>>>     remote commands in parallel.
>>>>>>>>>      > Score: Tiden: 2, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Logs control
>>>>>>>>>      > Both frameworks have similar builtin support for remote
>>>>>>>>> logs
>>>>>>>>>     collection and grepping. Tiden has built-in plugin that can
>>>>>>>>> zip,
>>>>>>>>>     collect arbitrary log files from arbitrary locations at
>>>>>>>>>     test/module/suite granularity and unzip if needed, also
>>> application
>>>>>>>>>     API to search / wait for messages in logs. Ducktape allows
>>>>>>>>> each
>>>>>>>>>     service declare its log files location (seemingly does not
>>> support
>>>>>>>>>     logs rollback), and a single entrypoint to collect service
>>>>>>>>> logs.
>>>>>>>>>      > Score: Tiden: 1, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Test assertions
>>>>>>>>>      > Tiden: simple asserts, also few customized assertion
>>>>>>>>> helpers.
>>>>>>>>>      > Ducktape: simple asserts.
>>>>>>>>>      > Score: Tiden: 2, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Test reporting
>>>>>>>>>      > Ducktape: limited to its own text/html format
>>>>>>>>>      > Tiden: provides text report, yaml report for reporting
>>>>>>>>> tools
>>>>>>>>>     integration, XML xUnit report for integration with
>>>>>> Jenkins/TeamCity.
>>>>>>>>>      > Score: Tiden: 3, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Provisioning and deployment
>>>>>>>>>      > Ducktape: can provision subset of hosts from cluster for
>>>>>>>>> test
>>>>>>>>>     needs. However, that means, that test can’t be scaled without
>>> test
>>>>>>>>>     code changes. Does not do any deploy, relies on external
>>>>>>>>> means,
>>>>>> e.g.
>>>>>>>>>     pre-packaged in docker image, as in PoC.
>>>>>>>>>      > Tiden: Given a set of hosts, Tiden uses all of them for the
>>>>>> test.
>>>>>>>>>     Provisioning should be done by external means. However,
>>>>>>>>> provides
>>> a
>>>>>>>>>     conventional automated deployment routines.
>>>>>>>>>      > Score: Tiden: 1, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Documentation and Extensibility
>>>>>>>>>      > Tiden: current API documentation is limited, should
>>>>>>>>> change as
>>> we
>>>>>>>>>     go open source. Tiden is easily extensible via hooks and
>>>>>>>>> plugins,
>>>>>>>>>     see example Maven plugin and Gatling application at [11].
>>>>>>>>>      > Ducktape: basic documentation at readthedocs.io
>>>>>>>>>     <http://readthedocs.io>. Codebase is rigid, framework core is
>>>>>>>>>     tightly coupled and hard to change. The only possible
>>>>>>>>> extension
>>>>>>>>>     mechanism is fork-and-rewrite.
>>>>>>>>>      > Score: Tiden: 2, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > I can continue more on this, but it should be enough for
>>>>>>>>> now:
>>>>>>>>>      > Overall score: Tiden: 22, Ducktape: 14.
>>>>>>>>>      >
>>>>>>>>>      > Time for discussion!
>>>>>>>>>      >
>>>>>>>>>      > ---
>>>>>>>>>      > [1] - https://www.testcontainers.org/
>>>>>>>>>      > [2] - http://arquillian.org/guides/getting_started/
>>>>>>>>>      > [3] - https://jmeter.apache.org/index.html
>>>>>>>>>      > [4] - https://openjdk.java.net/projects/code-tools/jmh/
>>>>>>>>>      > [5] - https://gatling.io/docs/current/
>>>>>>>>>      > [6] - https://github.com/gridgain/yardstick
>>>>>>>>>      > [7] - https://github.com/gridgain/poc-tester
>>>>>>>>>      > [8] -
>>>>>>>>>
>>>>>>
>>> https://cwiki.apache.org/confluence/display/KAFKA/System+Test+Improvements 
>>>
>>>>>>>>>      > [9] - https://github.com/gridgain/tiden
>>>>>>>>>      > [10] - https://pypi.org/project/jenkins-job-builder/
>>>>>>>>>      > [11] - https://github.com/mshonichev/tiden_examples
>>>>>>>>>      >
>>>>>>>>>      > On 25.05.2020 11:09, Nikolay Izhikov wrote:
>>>>>>>>>      >> Hello,
>>>>>>>>>      >>
>>>>>>>>>      >> Branch with duck tape created -
>>>>>>>>>     https://github.com/apache/ignite/tree/ignite-ducktape
>>>>>>>>>      >>
>>>>>>>>>      >> Any who are willing to contribute to PoC are welcome.
>>>>>>>>>      >>
>>>>>>>>>      >>
>>>>>>>>>      >>> 21 мая 2020 г., в 22:33, Nikolay Izhikov
>>>>>>>>>     <[hidden email] <mailto:[hidden email]>>
>>>>>> написал(а):
>>>>>>>>>      >>>
>>>>>>>>>      >>> Hello, Denis.
>>>>>>>>>      >>>
>>>>>>>>>      >>> There is no rush with these improvements.
>>>>>>>>>      >>> We can wait for Maxim proposal and compare two
>>>>>>>>> solutions :)
>>>>>>>>>      >>>
>>>>>>>>>      >>>> 21 мая 2020 г., в 22:24, Denis Magda <[hidden email]
>>>>>>>>>     <mailto:[hidden email]>> написал(а):
>>>>>>>>>      >>>>
>>>>>>>>>      >>>> Hi Nikolay,
>>>>>>>>>      >>>>
>>>>>>>>>      >>>> Thanks for kicking off this conversation and sharing
>>>>>>>>> your
>>>>>>>>>     findings with the
>>>>>>>>>      >>>> results. That's the right initiative. I do agree that
>>> Ignite
>>>>>>>>>     needs to have
>>>>>>>>>      >>>> an integration testing framework with capabilities
>>>>>>>>> listed
>>> by
>>>>>> you.
>>>>>>>>>      >>>>
>>>>>>>>>      >>>> As we discussed privately, I would only check if
>>>>>>>>> instead of
>>>>>>>>>      >>>> Confluent's Ducktape library, we can use an integration
>>>>>>>>>     testing framework
>>>>>>>>>      >>>> developed by GridGain for testing of Ignite/GridGain
>>>>>> clusters.
>>>>>>>>>     That
>>>>>>>>>      >>>> framework has been battle-tested and might be more
>>>>>> convenient for
>>>>>>>>>      >>>> Ignite-specific workloads. Let's wait for @Maksim
>>>>>>>>> Shonichev
>>>>>>>>>      >>>> <[hidden email]
>>>>>>>>> <mailto:[hidden email]>>
>>>>>> who
>>>>>>>>>     promised to join this thread once he finishes
>>>>>>>>>      >>>> preparing the usage examples of the framework. To my
>>>>>>>>>     knowledge, Max has
>>>>>>>>>      >>>> already been working on that for several days.
>>>>>>>>>      >>>>
>>>>>>>>>      >>>> -
>>>>>>>>>      >>>> Denis
>>>>>>>>>      >>>>
>>>>>>>>>      >>>>
>>>>>>>>>      >>>> On Thu, May 21, 2020 at 12:27 AM Nikolay Izhikov
>>>>>>>>>     <[hidden email] <mailto:[hidden email]>>
>>>>>>>>>      >>>> wrote:
>>>>>>>>>      >>>>
>>>>>>>>>      >>>>> Hello, Igniters.
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> I created a PoC [1] for the integration tests of
>>>>>>>>> Ignite.
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> Let me briefly explain the gap I want to cover:
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> 1. For now, we don’t have a solution for automated
>>>>>>>>> testing
>>>>>> of
>>>>>>>>>     Ignite on
>>>>>>>>>      >>>>> «real cluster».
>>>>>>>>>      >>>>> By «real cluster» I mean cluster «like a production»:
>>>>>>>>>      >>>>>       * client and server nodes deployed on different
>>> hosts.
>>>>>>>>>      >>>>>       * thin clients perform queries from some other
>>>>>>>>> hosts
>>>>>>>>>      >>>>>       * etc.
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> 2. We don’t have a solution for automated benchmarks of
>>> some
>>>>>>>>>     internal
>>>>>>>>>      >>>>> Ignite process
>>>>>>>>>      >>>>>       * PME
>>>>>>>>>      >>>>>       * rebalance.
>>>>>>>>>      >>>>> This means we don’t know - Do we perform
>>>>>>>>> rebalance(or PME)
>>>>>> in
>>>>>>>>>     2.7.0 faster
>>>>>>>>>      >>>>> or slower than in 2.8.0 for the same cluster?
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> 3. We don’t have a solution for automated testing of
>>> Ignite
>>>>>>>>>     integration in
>>>>>>>>>      >>>>> a real-world environment:
>>>>>>>>>      >>>>> Ignite-Spark integration can be taken as an example.
>>>>>>>>>      >>>>> I think some ML solutions also should be tested in
>>>>>> real-world
>>>>>>>>>     deployments.
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> Solution:
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> I propose to use duck tape library from confluent
>>>>>>>>> (apache
>>>>>> 2.0
>>>>>>>>>     license)
>>>>>>>>>      >>>>> I tested it both on the real cluster(Yandex Cloud)
>>>>>>>>> and on
>>>>>> the
>>>>>>>>>     local
>>>>>>>>>      >>>>> environment(docker) and it works just fine.
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> PoC contains following services:
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>>       * Simple rebalance test:
>>>>>>>>>      >>>>>               Start 2 server nodes,
>>>>>>>>>      >>>>>               Create some data with Ignite client,
>>>>>>>>>      >>>>>               Start one more server node,
>>>>>>>>>      >>>>>               Wait for rebalance finish
>>>>>>>>>      >>>>>       * Simple Ignite-Spark integration test:
>>>>>>>>>      >>>>>               Start 1 Spark master, start 1 Spark
>>>>>>>>> worker,
>>>>>>>>>      >>>>>               Start 1 Ignite server node
>>>>>>>>>      >>>>>               Create some data with Ignite client,
>>>>>>>>>      >>>>>               Check data in application that queries it
>>> from
>>>>>>>>>     Spark.
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> All tests are fully automated.
>>>>>>>>>      >>>>> Logs collection works just fine.
>>>>>>>>>      >>>>> You can see an example of the tests report - [4].
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> Pros:
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> * Ability to test local changes(no need to public
>>>>>>>>> changes
>>> to
>>>>>>>>>     some remote
>>>>>>>>>      >>>>> repository or similar).
>>>>>>>>>      >>>>> * Ability to parametrize test environment(run the same
>>> tests
>>>>>>>>>     on different
>>>>>>>>>      >>>>> JDK, JVM params, config, etc.)
>>>>>>>>>      >>>>> * Isolation by default so system tests are as
>>>>>>>>> reliable as
>>>>>>>>>     possible.
>>>>>>>>>      >>>>> * Utilities for pulling up and tearing down services
>>> easily
>>>>>>>>>     in clusters in
>>>>>>>>>      >>>>> different environments (e.g. local, custom cluster,
>>> Vagrant,
>>>>>>>>>     K8s, Mesos,
>>>>>>>>>      >>>>> Docker, cloud providers, etc.)
>>>>>>>>>      >>>>> * Easy to write unit tests for distributed systems
>>>>>>>>>      >>>>> * Adopted and successfully used by other distributed
>>>>>>>>> open
>>>>>>>>>     source project -
>>>>>>>>>      >>>>> Apache Kafka.
>>>>>>>>>      >>>>> * Collect results (e.g. logs, console output)
>>>>>>>>>      >>>>> * Report results (e.g. expected conditions met,
>>> performance
>>>>>>>>>     results, etc.)
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> WDYT?
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> [1] https://github.com/nizhikov/ignite/pull/15
>>>>>>>>>      >>>>> [2] https://github.com/confluentinc/ducktape
>>>>>>>>>      >>>>> [3]
>>>>>> https://ducktape-docs.readthedocs.io/en/latest/run_tests.html
>>>>>>>>>      >>>>> [4] https://yadi.sk/d/JC8ciJZjrkdndg
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>> <2020-07-05--004.tar.gz>
>>>
>>>
>>>
>>
>

[Max Shonichev]

Anton, Nikolay,

I want to share some more findings about ducktests I've stubmled upon
during porting them to Tiden.


First problem was that GridGain Tiden-based tests by default use real
production-like configuration for Ignite nodes, notably:

  - persitence enabled
  - ~120 caches in ~40 groups
  - data set size around 1M keys per each cache
  - primitive and PoJo cache values
  - extensive use of query entities (indices)

When I've tried to run 4 nodes with such configuration in docker, my
notebook nearly burns. Nevertheless, grid was starting and working OK,
but for one little 'but': each successive version under test was
starting slower and slower.

The 2.7.6 was the fastest, 2.8.0 and 2.8.1 a little bit slower, and your
fork (2.9.0-SNAPSHOT) failed to start 4 persistence-enabled nodes within
default 120 seconds timeout. In order to mimick behavior of your tests I
had to turn off persistence and use only 1 cache too.

It's a pity that you completely ignore persistence and indices in your
ducktests, otherwise you would quickly stuck into same limitation.

I hope in the nearest time I would adopt Tiden docker PoC to our
TeamCity and we'll try to git-bisect in order to find where this
slowdown comes from. After that I'll file a bug to IGNITE Jira.



Another problem with your rebalance benchmark is it's low accuracy due
to granularity of measurements.

You don't actually measure rebalance time, you measure time that takes
you to find a specific string in logs, that's confusing.

The scenario of your test is as follows:

1. start 3 server nodes
2. start 1 data loading client, preload a data, stop client
3. start 1 more server node
4. wait till server joins topology
5. wait till this server node completes exchange and write
'rebalanced=true, wasRebalanced=false' message to log
6. report time was taken by step 5 as 'Rebalance time'

Confusing thing here is that 'wait till' implementation - you actually
continuously re-scan logs sleeping each second and wait till message
appear. So, that means that rebalance time is at least of second
granularity or even higher, though it is reported with nanosecond
precision.

But for such lightweight configuration (single in-memory cache) and such
small set of data (1M keys only), rebalancing is very fast, and usually
performs under 1 second or just slightly slower.

Before waiting for rebalance message you first wait for topology message
and that wait also takes time to execute.

So, at the time Python part of the test performs first scan of the logs,
rebalancing is in most cases already done and time you report as
'0.0760810375213623' is actually the time to execute logs scanning code.

However, if rebalancing perform just a little bit slower after topology
update, then first scan of logs is failed, you sleep for whole one
second and rescan logs and there you got your message and report it as
'1.02205491065979'.

Under different conditions, dockerized application may run a little
slower or a little faster, that depends on overall system load, free
memory, etc. I've tried to increase load on my laptop by running browser
or maven build, and time to scan logs may fluctuate from 0.02 to 0.09 or
even 1.02 seconds. Note, that in CI environment, high system load from
tenants is a quite ordinary situation.

Suppose we adopted rebalance improvements and all versions after 2.9.0
would perform within 1 second just as 2.9.0 itself. Then your benchmark
would either report false negative (e.g. 0.02 for master and 0.03 for
PR), while actually on next re-run it would pass (e.g. 0.07 for master
and 0.03 for PR). That's not quite the 'stable and non-flacky' test
Ignite community wants.

What suggestions do you have to improve benchmark measurement accuracy?


A third question is about PME free switch benchmark. Under some
conditions, LongTxStreamerApplication actually hangs up PME. It need to
be investigated further, but either this was due to persistence enabled
or due to missing -DIGNITE_ALLOW_ATOMIC_OPS_IN_TX=false

Can you share some details about IGNITE_ALLOW_ATOMIC_OPS_IN_TX option?
Also, have you had performed a test of PME free switch with
persistence-enabled caches?


On 09.07.2020 10:11, Max Shonichev wrote:

> Anton,
>
> well, strange thing, but clean up and rerun helped.
>
>
> Ubuntu 18.04
>
> ====================================================================================================
>
> SESSION REPORT (ALL TESTS)
> ducktape version: 0.7.7
> session_id:       2020-07-06--003
> run time:         4 minutes 44.835 seconds
> tests run:        5
> passed:           5
> failed:           0
> ignored:          0
> ====================================================================================================
>
> test_id:
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=2.8.1
>
> status:     PASS
> run time:   41.927 seconds
> {"Rebalanced in (sec)": 1.02205491065979}
> ----------------------------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=dev
>
> status:     PASS
> run time:   51.985 seconds
> {"Rebalanced in (sec)": 0.0760810375213623}
> ----------------------------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=2.7.6
>
> status:     PASS
> run time:   1 minute 4.283 seconds
> {"Streamed txs": "1900", "Measure duration (ms)": "34818", "Worst
> latency (ms)": "31035"}
> ----------------------------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=dev
>
> status:     PASS
> run time:   1 minute 13.089 seconds
> {"Streamed txs": "73134", "Measure duration (ms)": "35843", "Worst
> latency (ms)": "139"}
> ----------------------------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.spark_integration_test.SparkIntegrationTest.test_spark_client
>
> status:     PASS
> run time:   53.332 seconds
> ----------------------------------------------------------------------------------------------------
>
>
>
> MacBook
> ================================================================================
>
> SESSION REPORT (ALL TESTS)
> ducktape version: 0.7.7
> session_id:       2020-07-06--001
> run time:         6 minutes 58.612 seconds
> tests run:        5
> passed:           5
> failed:           0
> ignored:          0
> ================================================================================
>
> test_id:
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=2.8.1
>
> status:     PASS
> run time:   48.724 seconds
> {"Rebalanced in (sec)": 3.2574470043182373}
> --------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=dev
>
> status:     PASS
> run time:   1 minute 23.210 seconds
> {"Rebalanced in (sec)": 2.165921211242676}
> --------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=2.7.6
>
> status:     PASS
> run time:   1 minute 12.659 seconds
> {"Streamed txs": "642", "Measure duration (ms)": "33177", "Worst latency
> (ms)": "31063"}
> --------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=dev
>
> status:     PASS
> run time:   1 minute 57.257 seconds
> {"Streamed txs": "32924", "Measure duration (ms)": "48252", "Worst
> latency (ms)": "1010"}
> --------------------------------------------------------------------------------
>
> test_id:
> ignitetest.tests.spark_integration_test.SparkIntegrationTest.test_spark_client
>
> status:     PASS
> run time:   1 minute 36.317 seconds
>
> =============
>
> while relative numbers proportion remains the same for different Ignite
> versions, absolute number for mac/linux differ more then twice.
>
> I'm finalizing code with 'local Tiden' appliance for your tests.  PR
> would be ready soon.
>
> Have you had a chance to deploy ducktests in bare metal?
>
>
>
> On 06.07.2020 14:27, Anton Vinogradov wrote:
>> Max,
>>
>> Thanks for the check!
>>
>>> Is it OK for those tests to fail?
>> No.
>> I see really strange things at logs.
>> Looks like you have concurrent ducktests run started not expected
>> services,
>> and this broke the tests.
>> Could you please clean up the docker (use clean-up script [1]).
>> Compile sources (use script [2]) and rerun the tests.
>>
>> [1]
>> https://github.com/anton-vinogradov/ignite/blob/dc98ee9df90b25eb5d928090b0e78b48cae2392e/modules/ducktests/tests/docker/clean_up.sh 
>>
>> [2]
>> https://github.com/anton-vinogradov/ignite/blob/3c39983005bd9eaf8cb458950d942fb592fff85c/scripts/build.sh 
>>
>>
>> On Mon, Jul 6, 2020 at 12:03 PM Nikolay Izhikov <[hidden email]>
>> wrote:
>>
>>> Hello, Maxim.
>>>
>>> Thanks for writing down the minutes.
>>>
>>> There is no such thing as «Nikolay team» on the dev-list.
>>> I propose to focus on product requirements and what we want to gain from
>>> the framework instead of taking into account the needs of some team.
>>>
>>> Can you, please, write down your version of requirements so we can
>>> reach a
>>> consensus on that and therefore move to the discussion of the
>>> implementation?
>>>
>>>> 6 июля 2020 г., в 11:18, Max Shonichev <[hidden email]> написал(а):
>>>>
>>>> Yes, Denis,
>>>>
>>>> common ground seems to be as follows:
>>>> Anton Vinogradov and Nikolay Izhikov would try to prepare and run PoC
>>> over physical hosts and share benchmark results. In the meantime,
>>> while I
>>> strongly believe that dockerized approach to benchmarking is a road to
>>> misleading and false positives, I'll prepare a PoC of Tiden in
>>> dockerized
>>> environment to support 'fast development prototyping' usecase Nikolay
>>> team
>>> insist on. It should be a matter of few days.
>>>>
>>>> As a side note, I've run Anton PoC locally and would like to have some
>>> comments about results:
>>>>
>>>> Test system: Ubuntu 18.04, docker 19.03.6
>>>> Test commands:
>>>>
>>>>
>>>> git clone -b ignite-ducktape [hidden email]:anton-vinogradov/ignite.git
>>>> cd ignite
>>>> mvn clean install -DskipTests -Dmaven.javadoc.skip=true
>>> -Pall-java,licenses,lgpl,examples,!spark-2.4,!spark,!scala
>>>> cd modules/ducktests/tests/docker
>>>> ./run_tests.sh
>>>>
>>>> Test results:
>>>>
>>> ====================================================================================================
>>>
>>>> SESSION REPORT (ALL TESTS)
>>>> ducktape version: 0.7.7
>>>> session_id:       2020-07-05--004
>>>> run time:         7 minutes 36.360 seconds
>>>> tests run:        5
>>>> passed:           3
>>>> failed:           2
>>>> ignored:          0
>>>>
>>> ====================================================================================================
>>>
>>>> test_id:
>>> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=2.8.1
>>>
>>>> status:     FAIL
>>>> run time:   3 minutes 12.232 seconds
>>>>
>>> ----------------------------------------------------------------------------------------------------
>>>
>>>> test_id:
>>> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=2.7.6
>>>
>>>> status:     FAIL
>>>> run time:   1 minute 33.076 seconds
>>>>
>>>>
>>>> Is it OK for those tests to fail? Attached is full test report
>>>>
>>>>
>>>> On 02.07.2020 17:46, Denis Magda wrote:
>>>>> Folks,
>>>>> Please share the summary of that Slack conversation here for records
>>> once
>>>>> you find common ground.
>>>>> -
>>>>> Denis
>>>>> On Thu, Jul 2, 2020 at 3:22 AM Nikolay Izhikov <[hidden email]>
>>> wrote:
>>>>>> Igniters.
>>>>>>
>>>>>> All who are interested in integration testing framework discussion
>>>>>> are
>>>>>> welcome into slack channel -
>>>>>>
>>> https://join.slack.com/share/zt-fk2ovehf-TcomEAwiXaPzLyNKZbmfzw?cdn_fallback=2 
>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>> 2 июля 2020 г., в 13:06, Anton Vinogradov <[hidden email]>
>>>>>>> написал(а):
>>>>>>>
>>>>>>> Max,
>>>>>>> Thanks for joining us.
>>>>>>>
>>>>>>>> 1. tiden can deploy artifacts by itself, while ducktape relies on
>>>>>>>> dependencies being deployed by external scripts.
>>>>>>> No. It is important to distinguish development, deploy, and
>>>>>> orchestration.
>>>>>>> All-in-one solutions have extremely limited usability.
>>>>>>> As to Ducktests:
>>>>>>> Docker is responsible for deployments during development.
>>>>>>> CI/CD is responsible for deployments during release and nightly
>>> checks.
>>>>>> It's up to the team to chose AWS, VM, BareMetal, and even OS.
>>>>>>> Ducktape is responsible for orchestration.
>>>>>>>
>>>>>>>> 2. tiden can execute actions over remote nodes in real parallel
>>>>>> fashion,
>>>>>>>> while ducktape internally does all actions sequentially.
>>>>>>> No. Ducktape may start any service in parallel. See Pme-free
>>>>>>> benchmark
>>>>>> [1] for details.
>>>>>>>
>>>>>>>> if we used ducktape solution we would have to instead prepare some
>>>>>>>> deployment scripts to pre-initialize Sberbank hosts, for example,
>>> with
>>>>>>>> Ansible or Chef.
>>>>>>> Sure, because a way of deploy depends on infrastructure.
>>>>>>> How can we be sure that OS we use and the restrictions we have
>>>>>>> will be
>>>>>> compatible with Tiden?
>>>>>>>
>>>>>>>> You have solved this deficiency with docker by putting all
>>> dependencies
>>>>>>>> into one uber-image ...
>>>>>>> and
>>>>>>>> I guess we all know about docker hyped ability to run over
>>> distributed
>>>>>>>> virtual networks.
>>>>>>> It is very important not to confuse the test's development (docker
>>> image
>>>>>> you're talking about) and real deployment.
>>>>>>>
>>>>>>>> If we had stopped and started 5 nodes one-by-one, as ducktape does
>>>>>>> All actions can be performed in parallel.
>>>>>>> See how Ducktests [2] starts cluster in parallel for example.
>>>>>>>
>>>>>>> [1]
>>>>>>
>>> https://github.com/apache/ignite/pull/7967/files#diff-59adde2a2ab7dc17aea6c65153dfcda7R84 
>>>
>>>>>>> [2]
>>>>>>
>>> https://github.com/apache/ignite/pull/7967/files#diff-d6a7b19f30f349d426b8894a40389cf5R79 
>>>
>>>>>>>
>>>>>>> On Thu, Jul 2, 2020 at 1:00 PM Nikolay Izhikov <[hidden email]>
>>>>>> wrote:
>>>>>>> Hello, Maxim.
>>>>>>>
>>>>>>>> 1. tiden can deploy artifacts by itself, while ducktape relies on
>>>>>> dependencies being deployed by external scripts
>>>>>>>
>>>>>>> Why do you think that maintaining deploy scripts coupled with the
>>>>>> testing framework is an advantage?
>>>>>>> I thought we want to see and maintain deployment scripts separate
>>>>>>> from
>>>>>> the testing framework.
>>>>>>>
>>>>>>>> 2. tiden can execute actions over remote nodes in real parallel
>>>>>> fashion, while ducktape internally does all actions sequentially.
>>>>>>>
>>>>>>> Can you, please, clarify, what actions do you have in mind?
>>>>>>> And why we want to execute them concurrently?
>>>>>>> Ignite node start, Client application execution can be done
>>> concurrently
>>>>>> with the ducktape approach.
>>>>>>>
>>>>>>>> If we used ducktape solution we would have to instead prepare some
>>>>>> deployment scripts to pre-initialize Sberbank hosts, for example,
>>>>>> with
>>>>>> Ansible or Chef
>>>>>>>
>>>>>>> We shouldn’t take some user approach as an argument in this
>>> discussion.
>>>>>> Let’s discuss a general approach for all users of the Ignite. Anyway,
>>> what
>>>>>> is wrong with the external deployment script approach?
>>>>>>>
>>>>>>> We, as a community, should provide several ways to run integration
>>> tests
>>>>>> out-of-the-box AND the ability to customize deployment regarding the
>>> user
>>>>>> landscape.
>>>>>>>
>>>>>>>> You have solved this deficiency with docker by putting all
>>>>>> dependencies into one uber-image and that looks like simple and
>>>>>> elegant
>>>>>> solution however, that effectively limits you to single-host testing.
>>>>>>>
>>>>>>> Docker image should be used only by the Ignite developers to test
>>>>>> something locally.
>>>>>>> It’s not intended for some real-world testing.
>>>>>>>
>>>>>>> The main issue with the Tiden that I see, it tested and
>>>>>>> maintained as
>>> a
>>>>>> closed source solution.
>>>>>>> This can lead to the hard to solve problems when we start using and
>>>>>> maintaining it as an open-source solution.
>>>>>>> Like, how many developers used Tiden? And how many of developers
>>>>>>> were
>>>>>> not authors of the Tiden itself?
>>>>>>>
>>>>>>>
>>>>>>>> 2 июля 2020 г., в 12:30, Max Shonichev <[hidden email]>
>>>>>> написал(а):
>>>>>>>>
>>>>>>>> Anton, Nikolay,
>>>>>>>>
>>>>>>>> Let's agree on what we are arguing about: whether it is about "like
>>> or
>>>>>> don't like" or about technical properties of suggested solutions.
>>>>>>>>
>>>>>>>> If it is about likes and dislikes, then the whole discussion is
>>>>>> meaningless. However, I hope together we can analyse pros and cons
>>>>>> carefully.
>>>>>>>>
>>>>>>>> As far as I can understand now, two main differences between
>>>>>>>> ducktape
>>>>>> and tiden is that:
>>>>>>>>
>>>>>>>> 1. tiden can deploy artifacts by itself, while ducktape relies on
>>>>>> dependencies being deployed by external scripts.
>>>>>>>>
>>>>>>>> 2. tiden can execute actions over remote nodes in real parallel
>>>>>> fashion, while ducktape internally does all actions sequentially.
>>>>>>>>
>>>>>>>> As for me, these are very important properties for distributed
>>> testing
>>>>>> framework.
>>>>>>>>
>>>>>>>> First property let us easily reuse tiden in existing
>>>>>>>> infrastructures,
>>>>>> for example, during Zookeeper IEP testing at Sberbank site we used
>>>>>> the
>>> same
>>>>>> tiden scripts that we use in our lab, the only change was putting a
>>> list of
>>>>>> hosts into config.
>>>>>>>>
>>>>>>>> If we used ducktape solution we would have to instead prepare some
>>>>>> deployment scripts to pre-initialize Sberbank hosts, for example,
>>>>>> with
>>>>>> Ansible or Chef.
>>>>>>>>
>>>>>>>>
>>>>>>>> You have solved this deficiency with docker by putting all
>>>>>> dependencies into one uber-image and that looks like simple and
>>>>>> elegant
>>>>>> solution,
>>>>>>>> however, that effectively limits you to single-host testing.
>>>>>>>>
>>>>>>>> I guess we all know about docker hyped ability to run over
>>> distributed
>>>>>> virtual networks. We used to go that way, but quickly found that
>>>>>> it is
>>> more
>>>>>> of the hype than real work. In real environments, there are problems
>>> with
>>>>>> routing, DNS, multicast and broadcast traffic, and many others, that
>>> turn
>>>>>> docker-based distributed solution into a fragile hard-to-maintain
>>> monster.
>>>>>>>>
>>>>>>>> Please, if you believe otherwise, perform a run of your PoC over at
>>>>>> least two physical hosts and share results with us.
>>>>>>>>
>>>>>>>> If you consider that one physical docker host is enough, please,
>>> don't
>>>>>> overlook that we want to run real scale scenarios, with 50-100 cache
>>>>>> groups, persistence enabled and a millions of keys loaded.
>>>>>>>>
>>>>>>>> Practical limit for such configurations is 4-6 nodes per single
>>>>>> physical host. Otherwise, tests become flaky due to resource
>>> starvation.
>>>>>>>>
>>>>>>>> Please, if you believe otherwise, perform at least a 10 of runs of
>>>>>> your PoC with other tests running at TC (we're targeting TeamCity,
>>> right?)
>>>>>> and share results so we could check if the numbers are reproducible.
>>>>>>>>
>>>>>>>> I stress this once more: functional integration tests are OK to run
>>> in
>>>>>> Docker and CI, but running benchmarks in Docker is a big NO GO.
>>>>>>>>
>>>>>>>>
>>>>>>>> Second property let us write tests that require real-parallel
>>>>>>>> actions
>>>>>> over hosts.
>>>>>>>>
>>>>>>>> For example, agreed scenario for PME benchmarkduring "PME
>>> optimization
>>>>>> stream" was as follows:
>>>>>>>>
>>>>>>>>   - 10 server nodes, preloaded with 1M of keys
>>>>>>>>   - 4 client nodes perform transactional load  (client nodes
>>> physically
>>>>>> separated from server nodes)
>>>>>>>>   - during load:
>>>>>>>>   -- 5 server nodes stopped in parallel
>>>>>>>>   -- after 1 minute, all 5 nodes are started in parallel
>>>>>>>>   - load stopped, logs are analysed for exchange times.
>>>>>>>>
>>>>>>>> If we had stopped and started 5 nodes one-by-one, as ducktape does,
>>>>>> then partition map exchange merge would not happen and we could not
>>> have
>>>>>> measured PME optimizations for that case.
>>>>>>>>
>>>>>>>>
>>>>>>>> These are limitations of ducktape that we believe as a more
>>>>>>>> important
>>>>>>>> argument "against" than you provide "for".
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> On 30.06.2020 14:58, Anton Vinogradov wrote:
>>>>>>>>> Folks,
>>>>>>>>> First, I've created PR [1] with ducktests improvements
>>>>>>>>> PR contains the following changes
>>>>>>>>> - Pme-free switch proof-benchmark (2.7.6 vs master)
>>>>>>>>> - Ability to check (compare with) previous releases (eg. 2.7.6 &
>>> 2.8)
>>>>>>>>> - Global refactoring
>>>>>>>>> -- benchmarks javacode simplification
>>>>>>>>> -- services python and java classes code deduplication
>>>>>>>>> -- fail-fast checks for java and python (eg. application should
>>>>>> explicitly write it finished with success)
>>>>>>>>> -- simple results extraction from tests and benchmarks
>>>>>>>>> -- javacode now configurable from tests/benchmarks
>>>>>>>>> -- proper SIGTERM handling at javacode (eg. it may finish last
>>>>>> operation and log results)
>>>>>>>>> -- docker volume now marked as delegated to increase execution
>>>>>>>>> speed
>>>>>> for mac & win users
>>>>>>>>> -- Ignite cluster now start in parallel (start speed-up)
>>>>>>>>> -- Ignite can be configured at test/benchmark
>>>>>>>>> - full and module assembly scripts added
>>>>>>>> Great job done! But let me remind one of Apache Ignite principles:
>>>>>>>> week of thinking save months of development.
>>>>>>>>
>>>>>>>>
>>>>>>>>> Second, I'd like to propose to accept ducktests [2] (ducktape
>>>>>> integration) as a target "PoC check & real topology benchmarking
>>>>>> tool".
>>>>>>>>> Ducktape pros
>>>>>>>>> - Developed for distributed system by distributed system
>>>>>>>>> developers.
>>>>>>>> So does Tiden
>>>>>>>>
>>>>>>>>> - Developed since 2014, stable.
>>>>>>>> Tiden is also pretty stable, and development start date is not a
>>>>>>>> good
>>>>>> argument, for example pytest is since 2004, pytest-xdist (plugin for
>>>>>> distributed testing) is since 2010, but we don't see it as a
>>> alternative at
>>>>>> all.
>>>>>>>>
>>>>>>>>> - Proven usability by usage at Kafka.
>>>>>>>> Tiden is proven usable by usage at GridGain and Sberbank
>>>>>>>> deployments.
>>>>>>>> Core, storage, sql and tx teams use benchmark results provided by
>>>>>> Tiden on a daily basis.
>>>>>>>>
>>>>>>>>> - Dozens of dozens tests and benchmarks at Kafka as a great
>>>>>>>>> example
>>>>>> pack.
>>>>>>>> We'll donate some of our suites to Ignite as I've mentioned in
>>>>>> previous letter.
>>>>>>>>
>>>>>>>>> - Built-in Docker support for rapid development and checks.
>>>>>>>> False, there's no specific 'docker support' in ducktape itself, you
>>>>>> just wrap it in docker by yourself, because ducktape is lacking
>>> deployment
>>>>>> abilities.
>>>>>>>>
>>>>>>>>> - Great for CI automation.
>>>>>>>> False, there's no specific CI-enabled features in ducktape.
>>>>>>>> Tiden, on
>>>>>> the other hand, provide generic xUnit reporting format, which is
>>> supported
>>>>>> by both TeamCity and Jenkins. Also, instead of using private keys,
>>> Tiden
>>>>>> can use SSH agent, which is also great for CI, because both
>>>>>>>> TeamCity and Jenkins store keys in secret storage available only
>>>>>>>> for
>>>>>> ssh-agent and only for the time of the test.
>>>>>>>>
>>>>>>>>
>>>>>>>>>> As an additional motivation, at least 3 teams
>>>>>>>>> - IEP-45 team (to check crash-recovery speed-up (discovery and
>>> Zabbix
>>>>>> speed-up))
>>>>>>>>> - Ignite SE Plugins team (to check plugin's features does not
>>>>>> slow-down or broke AI features)
>>>>>>>>> - Ignite SE QA team (to append already developed
>>>>>>>>> smoke/load/failover
>>>>>> tests to AI codebase)
>>>>>>>>
>>>>>>>> Please, before recommending your tests to other teams, provide
>>>>>>>> proofs
>>>>>>>> that your tests are reproducible in real environment.
>>>>>>>>
>>>>>>>>
>>>>>>>>> now, wait for ducktest merge to start checking cases they
>>>>>>>>> working on
>>>>>> in AI way.
>>>>>>>>> Thoughts?
>>>>>>>> Let us together review both solutions, we'll try to run your
>>>>>>>> tests in
>>>>>> our lab, and you'll try to at least checkout tiden and see if same
>>> tests
>>>>>> can be implemented with it?
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>> [1] https://github.com/apache/ignite/pull/7967
>>>>>>>>> [2] https://github.com/apache/ignite/tree/ignite-ducktape
>>>>>>>>> On Tue, Jun 16, 2020 at 12:22 PM Nikolay Izhikov <
>>> [hidden email]
>>>>>> <mailto:[hidden email]>> wrote:
>>>>>>>>>     Hello, Maxim.
>>>>>>>>>     Thank you for so detailed explanation.
>>>>>>>>>     Can we put the content of this discussion somewhere on the
>>>>>>>>> wiki?
>>>>>>>>>     So It doesn’t get lost.
>>>>>>>>>     I divide the answer in several parts. From the requirements to
>>> the
>>>>>>>>>     implementation.
>>>>>>>>>     So, if we agreed on the requirements we can proceed with the
>>>>>>>>>     discussion of the implementation.
>>>>>>>>>     1. Requirements:
>>>>>>>>>     The main goal I want to achieve is *reproducibility* of the
>>> tests.
>>>>>>>>>     I’m sick and tired with the zillions of flaky, rarely
>>>>>>>>> failed, and
>>>>>>>>>     almost never failed tests in Ignite codebase.
>>>>>>>>>     We should start with the simplest scenarios that will be as
>>>>>> reliable
>>>>>>>>>     as steel :)
>>>>>>>>>     I want to know for sure:
>>>>>>>>>        - Is this PR makes rebalance quicker or not?
>>>>>>>>>        - Is this PR makes PME quicker or not?
>>>>>>>>>     So, your description of the complex test scenario looks as
>>>>>>>>> a next
>>>>>>>>>     step to me.
>>>>>>>>>     Anyway, It’s cool we already have one.
>>>>>>>>>     The second goal is to have a strict test lifecycle as we
>>>>>>>>> have in
>>>>>>>>>     JUnit and similar frameworks.
>>>>>>>>>      > It covers production-like deployment and running a
>>>>>>>>> scenarios
>>>>>> over
>>>>>>>>>     a single database instance.
>>>>>>>>>     Do you mean «single cluster» or «single host»?
>>>>>>>>>     2. Existing tests:
>>>>>>>>>      > A Combinator suite allows to run set of operations
>>> concurrently
>>>>>>>>>     over given database instance.
>>>>>>>>>      > A Consumption suite allows to run a set production-like
>>> actions
>>>>>>>>>     over given set of Ignite/GridGain versions and compare test
>>> metrics
>>>>>>>>>     across versions
>>>>>>>>>      > A Yardstick suite
>>>>>>>>>      > A Stress suite that simulates hardware environment
>>>>>>>>> degradation
>>>>>>>>>      > An Ultimate, DR and Compatibility suites that performs
>>>>>> functional
>>>>>>>>>     regression testing
>>>>>>>>>      > Regression
>>>>>>>>>     Great news that we already have so many choices for testing!
>>>>>>>>>     Mature test base is a big +1 for Tiden.
>>>>>>>>>     3. Comparison:
>>>>>>>>>      > Criteria: Test configuration
>>>>>>>>>      > Ducktape: single JSON string for all tests
>>>>>>>>>      > Tiden: any number of YaML config files, command line option
>>> for
>>>>>>>>>     fine-grained test configuration, ability to select/modify
>>>>>>>>> tests
>>>>>>>>>     behavior based on Ignite version.
>>>>>>>>>     1. Many YAML files can be hard to maintain.
>>>>>>>>>     2. In ducktape, you can set parameters via «—parameters»
>>>>>>>>> option.
>>>>>>>>>     Please, take a look at the doc [1]
>>>>>>>>>      > Criteria: Cluster control
>>>>>>>>>      > Tiden: additionally can address cluster as a whole and
>>>>>>>>> execute
>>>>>>>>>     remote commands in parallel.
>>>>>>>>>     It seems we implement this ability in the PoC, already.
>>>>>>>>>      > Criteria: Test assertions
>>>>>>>>>      > Tiden: simple asserts, also few customized assertion
>>>>>>>>> helpers.
>>>>>>>>>      > Ducktape: simple asserts.
>>>>>>>>>     Can you, please, be more specific.
>>>>>>>>>     What helpers do you have in mind?
>>>>>>>>>     Ducktape has an asserts that waits for logfile messages or
>>>>>>>>> some
>>>>>>>>>     process finish.
>>>>>>>>>      > Criteria: Test reporting
>>>>>>>>>      > Ducktape: limited to its own text/HTML format
>>>>>>>>>     Ducktape have
>>>>>>>>>     1. Text reporter
>>>>>>>>>     2. Customizable HTML reporter
>>>>>>>>>     3. JSON reporter.
>>>>>>>>>     We can show JSON with the any template or tool.
>>>>>>>>>      > Criteria: Provisioning and deployment
>>>>>>>>>      > Ducktape: can provision subset of hosts from cluster for
>>>>>>>>> test
>>>>>>>>>     needs. However, that means, that test can’t be scaled without
>>> test
>>>>>>>>>     code changes. Does not do any deploy, relies on external
>>>>>>>>> means,
>>>>>> e.g.
>>>>>>>>>     pre-packaged in docker image, as in PoC.
>>>>>>>>>     This is not true.
>>>>>>>>>     1. We can set explicit test parameters(node number) via
>>> parameters.
>>>>>>>>>     We can increase client count of cluster size without test code
>>>>>> changes.
>>>>>>>>>     2. We have many choices for the test environment. These
>>>>>>>>> choices
>>> are
>>>>>>>>>     tested and used in other projects:
>>>>>>>>>              * docker
>>>>>>>>>              * vagrant
>>>>>>>>>              * private cloud(ssh access)
>>>>>>>>>              * ec2
>>>>>>>>>     Please, take a look at Kafka documentation [2]
>>>>>>>>>      > I can continue more on this, but it should be enough for
>>>>>>>>> now:
>>>>>>>>>     We need to go deeper! :)
>>>>>>>>>     [1]
>>>>>>>>>
>>>>>> https://ducktape-docs.readthedocs.io/en/latest/run_tests.html#options
>>>>>>>>>     [2]
>>>>>> https://github.com/apache/kafka/tree/trunk/tests#ec2-quickstart
>>>>>>>>>      > 9 июня 2020 г., в 17:25, Max A. Shonichev
>>>>>>>>> <[hidden email]
>>>>>>>>>     <mailto:[hidden email]>> написал(а):
>>>>>>>>>      >
>>>>>>>>>      > Greetings, Nikolay,
>>>>>>>>>      >
>>>>>>>>>      > First of all, thank you for you great effort preparing
>>>>>>>>> PoC of
>>>>>>>>>     integration testing to Ignite community.
>>>>>>>>>      >
>>>>>>>>>      > It’s a shame Ignite did not have at least some such
>>>>>>>>> tests yet,
>>>>>>>>>     however, GridGain, as a major contributor to Apache Ignite
>>>>>>>>> had a
>>>>>>>>>     profound collection of in-house tools to perform
>>>>>>>>> integration and
>>>>>>>>>     performance testing for years already and while we slowly
>>> consider
>>>>>>>>>     sharing our expertise with the community, your initiative
>>>>>>>>> makes
>>> us
>>>>>>>>>     drive that process a bit faster, thanks a lot!
>>>>>>>>>      >
>>>>>>>>>      > I reviewed your PoC and want to share a little about
>>>>>>>>> what we
>>> do
>>>>>>>>>     on our part, why and how, hope it would help community take
>>> proper
>>>>>>>>>     course.
>>>>>>>>>      >
>>>>>>>>>      > First I’ll do a brief overview of what decisions we made
>>>>>>>>> and
>>>>>> what
>>>>>>>>>     we do have in our private code base, next I’ll describe
>>>>>>>>> what we
>>>>>> have
>>>>>>>>>     already donated to the public and what we plan public next,
>>>>>>>>> then
>>>>>>>>>     I’ll compare both approaches highlighting deficiencies in
>>>>>>>>> order
>>> to
>>>>>>>>>     spur public discussion on the matter.
>>>>>>>>>      >
>>>>>>>>>      > It might seem strange to use Python to run Bash to run Java
>>>>>>>>>     applications because that introduces IT industry best of
>>>>>>>>> breed’ –
>>>>>>>>>     the Python dependency hell – to the Java application code
>>>>>>>>> base.
>>> The
>>>>>>>>>     only strangest decision one can made is to use Maven to run
>>> Docker
>>>>>>>>>     to run Bash to run Python to run Bash to run Java, but
>>>>>>>>> desperate
>>>>>>>>>     times call for desperate measures I guess.
>>>>>>>>>      >
>>>>>>>>>      > There are Java-based solutions for integration testing
>>>>>>>>> exists,
>>>>>>>>>     e.g. Testcontainers [1], Arquillian [2], etc, and they
>>>>>>>>> might go
>>>>>> well
>>>>>>>>>     for Ignite community CI pipelines by them selves. But we also
>>>>>> wanted
>>>>>>>>>     to run performance tests and benchmarks, like the dreaded PME
>>>>>>>>>     benchmark, and this is solved by totally different set of
>>>>>>>>> tools
>>> in
>>>>>>>>>     Java world, e.g. Jmeter [3], OpenJMH [4], Gatling [5], etc.
>>>>>>>>>      >
>>>>>>>>>      > Speaking specifically about benchmarking, Apache Ignite
>>>>>> community
>>>>>>>>>     already has Yardstick [6], and there’s nothing wrong with
>>>>>>>>> writing
>>>>>>>>>     PME benchmark using Yardstick, but we also wanted to be
>>>>>>>>> able to
>>> run
>>>>>>>>>     scenarios like this:
>>>>>>>>>      > - put an X load to a Ignite database;
>>>>>>>>>      > - perform an Y set of operations to check how Ignite copes
>>> with
>>>>>>>>>     operations under load.
>>>>>>>>>      >
>>>>>>>>>      > And yes, we also wanted applications under test be deployed
>>>>>> ‘like
>>>>>>>>>     in a production’, e.g. distributed over a set of hosts. This
>>> arises
>>>>>>>>>     questions about provisioning and nodes affinity which I’ll
>>>>>>>>> cover
>>> in
>>>>>>>>>     detail later.
>>>>>>>>>      >
>>>>>>>>>      > So we decided to put a little effort to build a simple
>>>>>>>>> tool to
>>>>>>>>>     cover different integration and performance scenarios, and
>>>>>>>>> our QA
>>>>>>>>>     lab first attempt was PoC-Tester [7], currently open source
>>>>>>>>> for
>>> all
>>>>>>>>>     but for reporting web UI. It’s a quite simple to use 95%
>>> Java-based
>>>>>>>>>     tool targeted to be run on a pre-release QA stage.
>>>>>>>>>      >
>>>>>>>>>      > It covers production-like deployment and running a
>>>>>>>>> scenarios
>>>>>> over
>>>>>>>>>     a single database instance. PoC-Tester scenarios consists of a
>>>>>>>>>     sequence of tasks running sequentially or in parallel.
>>>>>>>>> After all
>>>>>>>>>     tasks complete, or at any time during test, user can run logs
>>>>>>>>>     collection task, logs are checked against exceptions and a
>>> summary
>>>>>>>>>     of found issues and task ops/latency statistics is
>>>>>>>>> generated at
>>> the
>>>>>>>>>     end of scenario. One of the main PoC-Tester features is its
>>>>>>>>>     fire-and-forget approach to task managing. That is, you can
>>> deploy
>>>>>> a
>>>>>>>>>     grid and left it running for weeks, periodically firing some
>>> tasks
>>>>>>>>>     onto it.
>>>>>>>>>      >
>>>>>>>>>      > During earliest stages of PoC-Tester development it becomes
>>>>>> quite
>>>>>>>>>     clear that Java application development is a tedious
>>>>>>>>> process and
>>>>>>>>>     architecture decisions you take during development are slow
>>>>>>>>> and
>>>>>> hard
>>>>>>>>>     to change.
>>>>>>>>>      > For example, scenarios like this
>>>>>>>>>      > - deploy two instances of GridGain with master-slave data
>>>>>>>>>     replication configured;
>>>>>>>>>      > - put a load on master;
>>>>>>>>>      > - perform checks on slave,
>>>>>>>>>      > or like this:
>>>>>>>>>      > - preload a 1Tb of data by using your favorite tool of
>>>>>>>>> choice
>>> to
>>>>>>>>>     an Apache Ignite of version X;
>>>>>>>>>      > - run a set of functional tests running Apache Ignite
>>>>>>>>> version
>>> Y
>>>>>>>>>     over preloaded data,
>>>>>>>>>      > do not fit well in the PoC-Tester workflow.
>>>>>>>>>      >
>>>>>>>>>      > So, this is why we decided to use Python as a generic
>>> scripting
>>>>>>>>>     language of choice.
>>>>>>>>>      >
>>>>>>>>>      > Pros:
>>>>>>>>>      > - quicker prototyping and development cycles
>>>>>>>>>      > - easier to find DevOps/QA engineer with Python skills than
>>> one
>>>>>>>>>     with Java skills
>>>>>>>>>      > - used extensively all over the world for DevOps/CI
>>>>>>>>> pipelines
>>>>>> and
>>>>>>>>>     thus has rich set of libraries for all possible integration
>>>>>>>>> uses
>>>>>> cases.
>>>>>>>>>      >
>>>>>>>>>      > Cons:
>>>>>>>>>      > - Nightmare with dependencies. Better stick to specific
>>>>>>>>>     language/libraries version.
>>>>>>>>>      >
>>>>>>>>>      > Comparing alternatives for Python-based testing
>>>>>>>>> framework we
>>>>>> have
>>>>>>>>>     considered following requirements, somewhat similar to what
>>> you’ve
>>>>>>>>>     mentioned for Confluent [8] previously:
>>>>>>>>>      > - should be able run locally or distributed (bare metal
>>>>>>>>> or in
>>>>>> the
>>>>>>>>>     cloud)
>>>>>>>>>      > - should have built-in deployment facilities for
>>>>>>>>> applications
>>>>>>>>>     under test
>>>>>>>>>      > - should separate test configuration and test code
>>>>>>>>>      > -- be able to easily reconfigure tests by simple
>>>>>>>>> configuration
>>>>>>>>>     changes
>>>>>>>>>      > -- be able to easily scale test environment by simple
>>>>>>>>>     configuration changes
>>>>>>>>>      > -- be able to perform regression testing by simple
>>>>>>>>> switching
>>>>>>>>>     artifacts under test via configuration
>>>>>>>>>      > -- be able to run tests with different JDK version by
>>>>>>>>> simple
>>>>>>>>>     configuration changes
>>>>>>>>>      > - should have human readable reports and/or reporting tools
>>>>>>>>>     integration
>>>>>>>>>      > - should allow simple test progress monitoring, one does
>>>>>>>>> not
>>>>>> want
>>>>>>>>>     to run 6-hours test to find out that application actually
>>>>>>>>> crashed
>>>>>>>>>     during first hour.
>>>>>>>>>      > - should allow parallel execution of test actions
>>>>>>>>>      > - should have clean API for test writers
>>>>>>>>>      > -- clean API for distributed remote commands execution
>>>>>>>>>      > -- clean API for deployed applications start / stop and
>>>>>>>>> other
>>>>>>>>>     operations
>>>>>>>>>      > -- clean API for performing check on results
>>>>>>>>>      > - should be open source or at least source code should
>>>>>>>>> allow
>>>>>> ease
>>>>>>>>>     change or extension
>>>>>>>>>      >
>>>>>>>>>      > Back at that time we found no better alternative than to
>>>>>>>>> write
>>>>>>>>>     our own framework, and here goes Tiden [9] as GridGain
>>>>>>>>> framework
>>> of
>>>>>>>>>     choice for functional integration and performance testing.
>>>>>>>>>      >
>>>>>>>>>      > Pros:
>>>>>>>>>      > - solves all the requirements above
>>>>>>>>>      > Cons (for Ignite):
>>>>>>>>>      > - (currently) closed GridGain source
>>>>>>>>>      >
>>>>>>>>>      > On top of Tiden we’ve built a set of test suites, some of
>>> which
>>>>>>>>>     you might have heard already.
>>>>>>>>>      >
>>>>>>>>>      > A Combinator suite allows to run set of operations
>>> concurrently
>>>>>>>>>     over given database instance. Proven to find at least 30+ race
>>>>>>>>>     conditions and NPE issues.
>>>>>>>>>      >
>>>>>>>>>      > A Consumption suite allows to run a set production-like
>>> actions
>>>>>>>>>     over given set of Ignite/GridGain versions and compare test
>>> metrics
>>>>>>>>>     across versions, like heap/disk/CPU consumption, time to
>>>>>>>>> perform
>>>>>>>>>     actions, like client PME, server PME, rebalancing time, data
>>>>>>>>>     replication time, etc.
>>>>>>>>>      >
>>>>>>>>>      > A Yardstick suite is a thin layer of Python glue code to
>>>>>>>>> run
>>>>>>>>>     Apache Ignite pre-release benchmarks set. Yardstick itself
>>>>>>>>> has a
>>>>>>>>>     mediocre deployment capabilities, Tiden solves this easily.
>>>>>>>>>      >
>>>>>>>>>      > A Stress suite that simulates hardware environment
>>>>>>>>> degradation
>>>>>>>>>     during testing.
>>>>>>>>>      >
>>>>>>>>>      > An Ultimate, DR and Compatibility suites that performs
>>>>>> functional
>>>>>>>>>     regression testing of GridGain Ultimate Edition features like
>>>>>>>>>     snapshots, security, data replication, rolling upgrades, etc.
>>>>>>>>>      >
>>>>>>>>>      > A Regression and some IEPs testing suites, like IEP-14,
>>> IEP-15,
>>>>>>>>>     etc, etc, etc.
>>>>>>>>>      >
>>>>>>>>>      > Most of the suites above use another in-house developed
>>>>>>>>> Java
>>>>>> tool
>>>>>>>>>     – PiClient – to perform actual loading and miscellaneous
>>> operations
>>>>>>>>>     with Ignite under test. We use py4j Python-Java gateway
>>>>>>>>> library
>>> to
>>>>>>>>>     control PiClient instances from the tests.
>>>>>>>>>      >
>>>>>>>>>      > When we considered CI, we put TeamCity out of scope,
>>>>>>>>> because
>>>>>>>>>     distributed integration and performance tests tend to run for
>>> hours
>>>>>>>>>     and TeamCity agents are scarce and costly resource. So,
>>>>>>>>> bundled
>>>>>> with
>>>>>>>>>     Tiden there is jenkins-job-builder [10] based CI pipelines and
>>>>>>>>>     Jenkins xUnit reporting. Also, rich web UI tool Ward
>>>>>>>>> aggregates
>>>>>> test
>>>>>>>>>     run reports across versions and has built in visualization
>>> support
>>>>>>>>>     for Combinator suite.
>>>>>>>>>      >
>>>>>>>>>      > All of the above is currently closed source, but we plan to
>>> make
>>>>>>>>>     it public for community, and publishing Tiden core [9] is the
>>> first
>>>>>>>>>     step on that way. You can review some examples of using
>>>>>>>>> Tiden for
>>>>>>>>>     tests at my repository [11], for start.
>>>>>>>>>      >
>>>>>>>>>      > Now, let’s compare Ducktape PoC and Tiden.
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Language
>>>>>>>>>      > Tiden: Python, 3.7
>>>>>>>>>      > Ducktape: Python, proposes itself as Python 2.7, 3.6, 3.7
>>>>>>>>>     compatible, but actually can’t work with Python 3.7 due to
>>>>>>>>> broken
>>>>>>>>>     Zmq dependency.
>>>>>>>>>      > Comment: Python 3.7 has a much better support for
>>>>>>>>> async-style
>>>>>>>>>     code which might be crucial for distributed application
>>>>>>>>> testing.
>>>>>>>>>      > Score: Tiden: 1, Ducktape: 0
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Test writers API
>>>>>>>>>      > Supported integration test framework concepts are basically
>>> the
>>>>>> same:
>>>>>>>>>      > - a test controller (test runner)
>>>>>>>>>      > - a cluster
>>>>>>>>>      > - a node
>>>>>>>>>      > - an application (a service in Ducktape terms)
>>>>>>>>>      > - a test
>>>>>>>>>      > Score: Tiden: 5, Ducktape: 5
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Tests selection and run
>>>>>>>>>      > Ducktape: suite-package-class-method level selection,
>>>>>>>>> internal
>>>>>>>>>     scheduler allows to run tests in suite in parallel.
>>>>>>>>>      > Tiden: also suite-package-class-method level selection,
>>>>>>>>>     additionally allows selecting subset of tests by attribute,
>>>>>> parallel
>>>>>>>>>     runs not built in, but allows merging test reports after
>>> different
>>>>>> runs.
>>>>>>>>>      > Score: Tiden: 2, Ducktape: 2
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Test configuration
>>>>>>>>>      > Ducktape: single JSON string for all tests
>>>>>>>>>      > Tiden: any number of YaML config files, command line option
>>> for
>>>>>>>>>     fine-grained test configuration, ability to select/modify
>>>>>>>>> tests
>>>>>>>>>     behavior based on Ignite version.
>>>>>>>>>      > Score: Tiden: 3, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Cluster control
>>>>>>>>>      > Ducktape: allow execute remote commands by node granularity
>>>>>>>>>      > Tiden: additionally can address cluster as a whole and
>>>>>>>>> execute
>>>>>>>>>     remote commands in parallel.
>>>>>>>>>      > Score: Tiden: 2, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Logs control
>>>>>>>>>      > Both frameworks have similar builtin support for remote
>>>>>>>>> logs
>>>>>>>>>     collection and grepping. Tiden has built-in plugin that can
>>>>>>>>> zip,
>>>>>>>>>     collect arbitrary log files from arbitrary locations at
>>>>>>>>>     test/module/suite granularity and unzip if needed, also
>>> application
>>>>>>>>>     API to search / wait for messages in logs. Ducktape allows
>>>>>>>>> each
>>>>>>>>>     service declare its log files location (seemingly does not
>>> support
>>>>>>>>>     logs rollback), and a single entrypoint to collect service
>>>>>>>>> logs.
>>>>>>>>>      > Score: Tiden: 1, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Test assertions
>>>>>>>>>      > Tiden: simple asserts, also few customized assertion
>>>>>>>>> helpers.
>>>>>>>>>      > Ducktape: simple asserts.
>>>>>>>>>      > Score: Tiden: 2, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Test reporting
>>>>>>>>>      > Ducktape: limited to its own text/html format
>>>>>>>>>      > Tiden: provides text report, yaml report for reporting
>>>>>>>>> tools
>>>>>>>>>     integration, XML xUnit report for integration with
>>>>>> Jenkins/TeamCity.
>>>>>>>>>      > Score: Tiden: 3, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Provisioning and deployment
>>>>>>>>>      > Ducktape: can provision subset of hosts from cluster for
>>>>>>>>> test
>>>>>>>>>     needs. However, that means, that test can’t be scaled without
>>> test
>>>>>>>>>     code changes. Does not do any deploy, relies on external
>>>>>>>>> means,
>>>>>> e.g.
>>>>>>>>>     pre-packaged in docker image, as in PoC.
>>>>>>>>>      > Tiden: Given a set of hosts, Tiden uses all of them for the
>>>>>> test.
>>>>>>>>>     Provisioning should be done by external means. However,
>>>>>>>>> provides
>>> a
>>>>>>>>>     conventional automated deployment routines.
>>>>>>>>>      > Score: Tiden: 1, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > Criteria: Documentation and Extensibility
>>>>>>>>>      > Tiden: current API documentation is limited, should
>>>>>>>>> change as
>>> we
>>>>>>>>>     go open source. Tiden is easily extensible via hooks and
>>>>>>>>> plugins,
>>>>>>>>>     see example Maven plugin and Gatling application at [11].
>>>>>>>>>      > Ducktape: basic documentation at readthedocs.io
>>>>>>>>>     <http://readthedocs.io>. Codebase is rigid, framework core is
>>>>>>>>>     tightly coupled and hard to change. The only possible
>>>>>>>>> extension
>>>>>>>>>     mechanism is fork-and-rewrite.
>>>>>>>>>      > Score: Tiden: 2, Ducktape: 1
>>>>>>>>>      >
>>>>>>>>>      > I can continue more on this, but it should be enough for
>>>>>>>>> now:
>>>>>>>>>      > Overall score: Tiden: 22, Ducktape: 14.
>>>>>>>>>      >
>>>>>>>>>      > Time for discussion!
>>>>>>>>>      >
>>>>>>>>>      > ---
>>>>>>>>>      > [1] - https://www.testcontainers.org/
>>>>>>>>>      > [2] - http://arquillian.org/guides/getting_started/
>>>>>>>>>      > [3] - https://jmeter.apache.org/index.html
>>>>>>>>>      > [4] - https://openjdk.java.net/projects/code-tools/jmh/
>>>>>>>>>      > [5] - https://gatling.io/docs/current/
>>>>>>>>>      > [6] - https://github.com/gridgain/yardstick
>>>>>>>>>      > [7] - https://github.com/gridgain/poc-tester
>>>>>>>>>      > [8] -
>>>>>>>>>
>>>>>>
>>> https://cwiki.apache.org/confluence/display/KAFKA/System+Test+Improvements 
>>>
>>>>>>>>>      > [9] - https://github.com/gridgain/tiden
>>>>>>>>>      > [10] - https://pypi.org/project/jenkins-job-builder/
>>>>>>>>>      > [11] - https://github.com/mshonichev/tiden_examples
>>>>>>>>>      >
>>>>>>>>>      > On 25.05.2020 11:09, Nikolay Izhikov wrote:
>>>>>>>>>      >> Hello,
>>>>>>>>>      >>
>>>>>>>>>      >> Branch with duck tape created -
>>>>>>>>>     https://github.com/apache/ignite/tree/ignite-ducktape
>>>>>>>>>      >>
>>>>>>>>>      >> Any who are willing to contribute to PoC are welcome.
>>>>>>>>>      >>
>>>>>>>>>      >>
>>>>>>>>>      >>> 21 мая 2020 г., в 22:33, Nikolay Izhikov
>>>>>>>>>     <[hidden email] <mailto:[hidden email]>>
>>>>>> написал(а):
>>>>>>>>>      >>>
>>>>>>>>>      >>> Hello, Denis.
>>>>>>>>>      >>>
>>>>>>>>>      >>> There is no rush with these improvements.
>>>>>>>>>      >>> We can wait for Maxim proposal and compare two
>>>>>>>>> solutions :)
>>>>>>>>>      >>>
>>>>>>>>>      >>>> 21 мая 2020 г., в 22:24, Denis Magda <[hidden email]
>>>>>>>>>     <mailto:[hidden email]>> написал(а):
>>>>>>>>>      >>>>
>>>>>>>>>      >>>> Hi Nikolay,
>>>>>>>>>      >>>>
>>>>>>>>>      >>>> Thanks for kicking off this conversation and sharing
>>>>>>>>> your
>>>>>>>>>     findings with the
>>>>>>>>>      >>>> results. That's the right initiative. I do agree that
>>> Ignite
>>>>>>>>>     needs to have
>>>>>>>>>      >>>> an integration testing framework with capabilities
>>>>>>>>> listed
>>> by
>>>>>> you.
>>>>>>>>>      >>>>
>>>>>>>>>      >>>> As we discussed privately, I would only check if
>>>>>>>>> instead of
>>>>>>>>>      >>>> Confluent's Ducktape library, we can use an integration
>>>>>>>>>     testing framework
>>>>>>>>>      >>>> developed by GridGain for testing of Ignite/GridGain
>>>>>> clusters.
>>>>>>>>>     That
>>>>>>>>>      >>>> framework has been battle-tested and might be more
>>>>>> convenient for
>>>>>>>>>      >>>> Ignite-specific workloads. Let's wait for @Maksim
>>>>>>>>> Shonichev
>>>>>>>>>      >>>> <[hidden email]
>>>>>>>>> <mailto:[hidden email]>>
>>>>>> who
>>>>>>>>>     promised to join this thread once he finishes
>>>>>>>>>      >>>> preparing the usage examples of the framework. To my
>>>>>>>>>     knowledge, Max has
>>>>>>>>>      >>>> already been working on that for several days.
>>>>>>>>>      >>>>
>>>>>>>>>      >>>> -
>>>>>>>>>      >>>> Denis
>>>>>>>>>      >>>>
>>>>>>>>>      >>>>
>>>>>>>>>      >>>> On Thu, May 21, 2020 at 12:27 AM Nikolay Izhikov
>>>>>>>>>     <[hidden email] <mailto:[hidden email]>>
>>>>>>>>>      >>>> wrote:
>>>>>>>>>      >>>>
>>>>>>>>>      >>>>> Hello, Igniters.
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> I created a PoC [1] for the integration tests of
>>>>>>>>> Ignite.
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> Let me briefly explain the gap I want to cover:
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> 1. For now, we don’t have a solution for automated
>>>>>>>>> testing
>>>>>> of
>>>>>>>>>     Ignite on
>>>>>>>>>      >>>>> «real cluster».
>>>>>>>>>      >>>>> By «real cluster» I mean cluster «like a production»:
>>>>>>>>>      >>>>>       * client and server nodes deployed on different
>>> hosts.
>>>>>>>>>      >>>>>       * thin clients perform queries from some other
>>>>>>>>> hosts
>>>>>>>>>      >>>>>       * etc.
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> 2. We don’t have a solution for automated benchmarks of
>>> some
>>>>>>>>>     internal
>>>>>>>>>      >>>>> Ignite process
>>>>>>>>>      >>>>>       * PME
>>>>>>>>>      >>>>>       * rebalance.
>>>>>>>>>      >>>>> This means we don’t know - Do we perform
>>>>>>>>> rebalance(or PME)
>>>>>> in
>>>>>>>>>     2.7.0 faster
>>>>>>>>>      >>>>> or slower than in 2.8.0 for the same cluster?
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> 3. We don’t have a solution for automated testing of
>>> Ignite
>>>>>>>>>     integration in
>>>>>>>>>      >>>>> a real-world environment:
>>>>>>>>>      >>>>> Ignite-Spark integration can be taken as an example.
>>>>>>>>>      >>>>> I think some ML solutions also should be tested in
>>>>>> real-world
>>>>>>>>>     deployments.
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> Solution:
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> I propose to use duck tape library from confluent
>>>>>>>>> (apache
>>>>>> 2.0
>>>>>>>>>     license)
>>>>>>>>>      >>>>> I tested it both on the real cluster(Yandex Cloud)
>>>>>>>>> and on
>>>>>> the
>>>>>>>>>     local
>>>>>>>>>      >>>>> environment(docker) and it works just fine.
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> PoC contains following services:
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>>       * Simple rebalance test:
>>>>>>>>>      >>>>>               Start 2 server nodes,
>>>>>>>>>      >>>>>               Create some data with Ignite client,
>>>>>>>>>      >>>>>               Start one more server node,
>>>>>>>>>      >>>>>               Wait for rebalance finish
>>>>>>>>>      >>>>>       * Simple Ignite-Spark integration test:
>>>>>>>>>      >>>>>               Start 1 Spark master, start 1 Spark
>>>>>>>>> worker,
>>>>>>>>>      >>>>>               Start 1 Ignite server node
>>>>>>>>>      >>>>>               Create some data with Ignite client,
>>>>>>>>>      >>>>>               Check data in application that queries it
>>> from
>>>>>>>>>     Spark.
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> All tests are fully automated.
>>>>>>>>>      >>>>> Logs collection works just fine.
>>>>>>>>>      >>>>> You can see an example of the tests report - [4].
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> Pros:
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> * Ability to test local changes(no need to public
>>>>>>>>> changes
>>> to
>>>>>>>>>     some remote
>>>>>>>>>      >>>>> repository or similar).
>>>>>>>>>      >>>>> * Ability to parametrize test environment(run the same
>>> tests
>>>>>>>>>     on different
>>>>>>>>>      >>>>> JDK, JVM params, config, etc.)
>>>>>>>>>      >>>>> * Isolation by default so system tests are as
>>>>>>>>> reliable as
>>>>>>>>>     possible.
>>>>>>>>>      >>>>> * Utilities for pulling up and tearing down services
>>> easily
>>>>>>>>>     in clusters in
>>>>>>>>>      >>>>> different environments (e.g. local, custom cluster,
>>> Vagrant,
>>>>>>>>>     K8s, Mesos,
>>>>>>>>>      >>>>> Docker, cloud providers, etc.)
>>>>>>>>>      >>>>> * Easy to write unit tests for distributed systems
>>>>>>>>>      >>>>> * Adopted and successfully used by other distributed
>>>>>>>>> open
>>>>>>>>>     source project -
>>>>>>>>>      >>>>> Apache Kafka.
>>>>>>>>>      >>>>> * Collect results (e.g. logs, console output)
>>>>>>>>>      >>>>> * Report results (e.g. expected conditions met,
>>> performance
>>>>>>>>>     results, etc.)
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> WDYT?
>>>>>>>>>      >>>>>
>>>>>>>>>      >>>>> [1] https://github.com/nizhikov/ignite/pull/15
>>>>>>>>>      >>>>> [2] https://github.com/confluentinc/ducktape
>>>>>>>>>      >>>>> [3]
>>>>>> https://ducktape-docs.readthedocs.io/en/latest/run_tests.html
>>>>>>>>>      >>>>> [4] https://yadi.sk/d/JC8ciJZjrkdndg
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>> <2020-07-05--004.tar.gz>
>>>
>>>
>>>
>>
>

[Anton Vinogradov-2]

Discussed privately with Max.
Discussion results provided at the slack channel [1].

[1] https://the-asf.slack.com/archives/C016F4PS8KV/p1595336751234500

On Wed, Jul 15, 2020 at 3:59 PM Max Shonichev <[hidden email]> wrote:

> Anton, Nikolay,
>
> I want to share some more findings about ducktests I've stubmled upon
> during porting them to Tiden.
>
>
> First problem was that GridGain Tiden-based tests by default use real
> production-like configuration for Ignite nodes, notably:
>
>   - persitence enabled
>   - ~120 caches in ~40 groups
>   - data set size around 1M keys per each cache
>   - primitive and PoJo cache values
>   - extensive use of query entities (indices)
>
> When I've tried to run 4 nodes with such configuration in docker, my
> notebook nearly burns. Nevertheless, grid was starting and working OK,
> but for one little 'but': each successive version under test was
> starting slower and slower.
>
> The 2.7.6 was the fastest, 2.8.0 and 2.8.1 a little bit slower, and your
> fork (2.9.0-SNAPSHOT) failed to start 4 persistence-enabled nodes within
> default 120 seconds timeout. In order to mimick behavior of your tests I
> had to turn off persistence and use only 1 cache too.
>
> It's a pity that you completely ignore persistence and indices in your
> ducktests, otherwise you would quickly stuck into same limitation.
>
> I hope in the nearest time I would adopt Tiden docker PoC to our
> TeamCity and we'll try to git-bisect in order to find where this
> slowdown comes from. After that I'll file a bug to IGNITE Jira.
>
>
>
> Another problem with your rebalance benchmark is it's low accuracy due
> to granularity of measurements.
>
> You don't actually measure rebalance time, you measure time that takes
> you to find a specific string in logs, that's confusing.
>
> The scenario of your test is as follows:
>
> 1. start 3 server nodes
> 2. start 1 data loading client, preload a data, stop client
> 3. start 1 more server node
> 4. wait till server joins topology
> 5. wait till this server node completes exchange and write
> 'rebalanced=true, wasRebalanced=false' message to log
> 6. report time was taken by step 5 as 'Rebalance time'
>
> Confusing thing here is that 'wait till' implementation - you actually
> continuously re-scan logs sleeping each second and wait till message
> appear. So, that means that rebalance time is at least of second
> granularity or even higher, though it is reported with nanosecond
> precision.
>
> But for such lightweight configuration (single in-memory cache) and such
> small set of data (1M keys only), rebalancing is very fast, and usually
> performs under 1 second or just slightly slower.
>
> Before waiting for rebalance message you first wait for topology message
> and that wait also takes time to execute.
>
> So, at the time Python part of the test performs first scan of the logs,
> rebalancing is in most cases already done and time you report as
> '0.0760810375213623' is actually the time to execute logs scanning code.
>
> However, if rebalancing perform just a little bit slower after topology
> update, then first scan of logs is failed, you sleep for whole one
> second and rescan logs and there you got your message and report it as
> '1.02205491065979'.
>
> Under different conditions, dockerized application may run a little
> slower or a little faster, that depends on overall system load, free
> memory, etc. I've tried to increase load on my laptop by running browser
> or maven build, and time to scan logs may fluctuate from 0.02 to 0.09 or
> even 1.02 seconds. Note, that in CI environment, high system load from
> tenants is a quite ordinary situation.
>
> Suppose we adopted rebalance improvements and all versions after 2.9.0
> would perform within 1 second just as 2.9.0 itself. Then your benchmark
> would either report false negative (e.g. 0.02 for master and 0.03 for
> PR), while actually on next re-run it would pass (e.g. 0.07 for master
> and 0.03 for PR). That's not quite the 'stable and non-flacky' test
> Ignite community wants.
>
> What suggestions do you have to improve benchmark measurement accuracy?
>
>
> A third question is about PME free switch benchmark. Under some
> conditions, LongTxStreamerApplication actually hangs up PME. It need to
> be investigated further, but either this was due to persistence enabled
> or due to missing -DIGNITE_ALLOW_ATOMIC_OPS_IN_TX=false
>
> Can you share some details about IGNITE_ALLOW_ATOMIC_OPS_IN_TX option?
> Also, have you had performed a test of PME free switch with
> persistence-enabled caches?
>
>
> On 09.07.2020 10:11, Max Shonichev wrote:
> > Anton,
> >
> > well, strange thing, but clean up and rerun helped.
> >
> >
> > Ubuntu 18.04
> >
> >
> ====================================================================================================
>
> >
> > SESSION REPORT (ALL TESTS)
> > ducktape version: 0.7.7
> > session_id:       2020-07-06--003
> > run time:         4 minutes 44.835 seconds
> > tests run:        5
> > passed:           5
> > failed:           0
> > ignored:          0
> >
> ====================================================================================================
>
> >
> > test_id:
> >
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=2.8.1
>
> >
> > status:     PASS
> > run time:   41.927 seconds
> > {"Rebalanced in (sec)": 1.02205491065979}
> >
> ----------------------------------------------------------------------------------------------------
>
> >
> > test_id:
> >
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=dev
>
> >
> > status:     PASS
> > run time:   51.985 seconds
> > {"Rebalanced in (sec)": 0.0760810375213623}
> >
> ----------------------------------------------------------------------------------------------------
>
> >
> > test_id:
> >
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=2.7.6
>
> >
> > status:     PASS
> > run time:   1 minute 4.283 seconds
> > {"Streamed txs": "1900", "Measure duration (ms)": "34818", "Worst
> > latency (ms)": "31035"}
> >
> ----------------------------------------------------------------------------------------------------
>
> >
> > test_id:
> >
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=dev
>
> >
> > status:     PASS
> > run time:   1 minute 13.089 seconds
> > {"Streamed txs": "73134", "Measure duration (ms)": "35843", "Worst
> > latency (ms)": "139"}
> >
> ----------------------------------------------------------------------------------------------------
>
> >
> > test_id:
> >
> ignitetest.tests.spark_integration_test.SparkIntegrationTest.test_spark_client
>
> >
> > status:     PASS
> > run time:   53.332 seconds
> >
> ----------------------------------------------------------------------------------------------------
>
> >
> >
> >
> > MacBook
> >
> ================================================================================
>
> >
> > SESSION REPORT (ALL TESTS)
> > ducktape version: 0.7.7
> > session_id:       2020-07-06--001
> > run time:         6 minutes 58.612 seconds
> > tests run:        5
> > passed:           5
> > failed:           0
> > ignored:          0
> >
> ================================================================================
>
> >
> > test_id:
> >
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=2.8.1
>
> >
> > status:     PASS
> > run time:   48.724 seconds
> > {"Rebalanced in (sec)": 3.2574470043182373}
> >
> --------------------------------------------------------------------------------
>
> >
> > test_id:
> >
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=dev
>
> >
> > status:     PASS
> > run time:   1 minute 23.210 seconds
> > {"Rebalanced in (sec)": 2.165921211242676}
> >
> --------------------------------------------------------------------------------
>
> >
> > test_id:
> >
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=2.7.6
>
> >
> > status:     PASS
> > run time:   1 minute 12.659 seconds
> > {"Streamed txs": "642", "Measure duration (ms)": "33177", "Worst latency
> > (ms)": "31063"}
> >
> --------------------------------------------------------------------------------
>
> >
> > test_id:
> >
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=dev
>
> >
> > status:     PASS
> > run time:   1 minute 57.257 seconds
> > {"Streamed txs": "32924", "Measure duration (ms)": "48252", "Worst
> > latency (ms)": "1010"}
> >
> --------------------------------------------------------------------------------
>
> >
> > test_id:
> >
> ignitetest.tests.spark_integration_test.SparkIntegrationTest.test_spark_client
>
> >
> > status:     PASS
> > run time:   1 minute 36.317 seconds
> >
> > =============
> >
> > while relative numbers proportion remains the same for different Ignite
> > versions, absolute number for mac/linux differ more then twice.
> >
> > I'm finalizing code with 'local Tiden' appliance for your tests.  PR
> > would be ready soon.
> >
> > Have you had a chance to deploy ducktests in bare metal?
> >
> >
> >
> > On 06.07.2020 14:27, Anton Vinogradov wrote:
> >> Max,
> >>
> >> Thanks for the check!
> >>
> >>> Is it OK for those tests to fail?
> >> No.
> >> I see really strange things at logs.
> >> Looks like you have concurrent ducktests run started not expected
> >> services,
> >> and this broke the tests.
> >> Could you please clean up the docker (use clean-up script [1]).
> >> Compile sources (use script [2]) and rerun the tests.
> >>
> >> [1]
> >>
> https://github.com/anton-vinogradov/ignite/blob/dc98ee9df90b25eb5d928090b0e78b48cae2392e/modules/ducktests/tests/docker/clean_up.sh
> >>
> >> [2]
> >>
> https://github.com/anton-vinogradov/ignite/blob/3c39983005bd9eaf8cb458950d942fb592fff85c/scripts/build.sh
> >>
> >>
> >> On Mon, Jul 6, 2020 at 12:03 PM Nikolay Izhikov <[hidden email]>
> >> wrote:
> >>
> >>> Hello, Maxim.
> >>>
> >>> Thanks for writing down the minutes.
> >>>
> >>> There is no such thing as «Nikolay team» on the dev-list.
> >>> I propose to focus on product requirements and what we want to gain
> from
> >>> the framework instead of taking into account the needs of some team.
> >>>
> >>> Can you, please, write down your version of requirements so we can
> >>> reach a
> >>> consensus on that and therefore move to the discussion of the
> >>> implementation?
> >>>
> >>>> 6 июля 2020 г., в 11:18, Max Shonichev <[hidden email]>
> написал(а):
> >>>>
> >>>> Yes, Denis,
> >>>>
> >>>> common ground seems to be as follows:
> >>>> Anton Vinogradov and Nikolay Izhikov would try to prepare and run PoC
> >>> over physical hosts and share benchmark results. In the meantime,
> >>> while I
> >>> strongly believe that dockerized approach to benchmarking is a road to
> >>> misleading and false positives, I'll prepare a PoC of Tiden in
> >>> dockerized
> >>> environment to support 'fast development prototyping' usecase Nikolay
> >>> team
> >>> insist on. It should be a matter of few days.
> >>>>
> >>>> As a side note, I've run Anton PoC locally and would like to have some
> >>> comments about results:
> >>>>
> >>>> Test system: Ubuntu 18.04, docker 19.03.6
> >>>> Test commands:
> >>>>
> >>>>
> >>>> git clone -b ignite-ducktape [hidden email]:
> anton-vinogradov/ignite.git
> >>>> cd ignite
> >>>> mvn clean install -DskipTests -Dmaven.javadoc.skip=true
> >>> -Pall-java,licenses,lgpl,examples,!spark-2.4,!spark,!scala
> >>>> cd modules/ducktests/tests/docker
> >>>> ./run_tests.sh
> >>>>
> >>>> Test results:
> >>>>
> >>>
> ====================================================================================================
>
> >>>
> >>>> SESSION REPORT (ALL TESTS)
> >>>> ducktape version: 0.7.7
> >>>> session_id:       2020-07-05--004
> >>>> run time:         7 minutes 36.360 seconds
> >>>> tests run:        5
> >>>> passed:           3
> >>>> failed:           2
> >>>> ignored:          0
> >>>>
> >>>
> ====================================================================================================
>
> >>>
> >>>> test_id:
> >>>
> ignitetest.tests.benchmarks.add_node_rebalance_test.AddNodeRebalanceTest.test_add_node.version=2.8.1
>
> >>>
> >>>> status:     FAIL
> >>>> run time:   3 minutes 12.232 seconds
> >>>>
> >>>
> ----------------------------------------------------------------------------------------------------
>
> >>>
> >>>> test_id:
> >>>
> ignitetest.tests.benchmarks.pme_free_switch_test.PmeFreeSwitchTest.test.version=2.7.6
>
> >>>
> >>>> status:     FAIL
> >>>> run time:   1 minute 33.076 seconds
> >>>>
> >>>>
> >>>> Is it OK for those tests to fail? Attached is full test report
> >>>>
> >>>>
> >>>> On 02.07.2020 17:46, Denis Magda wrote:
> >>>>> Folks,
> >>>>> Please share the summary of that Slack conversation here for records
> >>> once
> >>>>> you find common ground.
> >>>>> -
> >>>>> Denis
> >>>>> On Thu, Jul 2, 2020 at 3:22 AM Nikolay Izhikov <[hidden email]>
> >>> wrote:
> >>>>>> Igniters.
> >>>>>>
> >>>>>> All who are interested in integration testing framework discussion
> >>>>>> are
> >>>>>> welcome into slack channel -
> >>>>>>
> >>>
> https://join.slack.com/share/zt-fk2ovehf-TcomEAwiXaPzLyNKZbmfzw?cdn_fallback=2
> >>>
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>>> 2 июля 2020 г., в 13:06, Anton Vinogradov <[hidden email]>
> >>>>>>> написал(а):
> >>>>>>>
> >>>>>>> Max,
> >>>>>>> Thanks for joining us.
> >>>>>>>
> >>>>>>>> 1. tiden can deploy artifacts by itself, while ducktape relies on
> >>>>>>>> dependencies being deployed by external scripts.
> >>>>>>> No. It is important to distinguish development, deploy, and
> >>>>>> orchestration.
> >>>>>>> All-in-one solutions have extremely limited usability.
> >>>>>>> As to Ducktests:
> >>>>>>> Docker is responsible for deployments during development.
> >>>>>>> CI/CD is responsible for deployments during release and nightly
> >>> checks.
> >>>>>> It's up to the team to chose AWS, VM, BareMetal, and even OS.
> >>>>>>> Ducktape is responsible for orchestration.
> >>>>>>>
> >>>>>>>> 2. tiden can execute actions over remote nodes in real parallel
> >>>>>> fashion,
> >>>>>>>> while ducktape internally does all actions sequentially.
> >>>>>>> No. Ducktape may start any service in parallel. See Pme-free
> >>>>>>> benchmark
> >>>>>> [1] for details.
> >>>>>>>
> >>>>>>>> if we used ducktape solution we would have to instead prepare some
> >>>>>>>> deployment scripts to pre-initialize Sberbank hosts, for example,
> >>> with
> >>>>>>>> Ansible or Chef.
> >>>>>>> Sure, because a way of deploy depends on infrastructure.
> >>>>>>> How can we be sure that OS we use and the restrictions we have
> >>>>>>> will be
> >>>>>> compatible with Tiden?
> >>>>>>>
> >>>>>>>> You have solved this deficiency with docker by putting all
> >>> dependencies
> >>>>>>>> into one uber-image ...
> >>>>>>> and
> >>>>>>>> I guess we all know about docker hyped ability to run over
> >>> distributed
> >>>>>>>> virtual networks.
> >>>>>>> It is very important not to confuse the test's development (docker
> >>> image
> >>>>>> you're talking about) and real deployment.
> >>>>>>>
> >>>>>>>> If we had stopped and started 5 nodes one-by-one, as ducktape does
> >>>>>>> All actions can be performed in parallel.
> >>>>>>> See how Ducktests [2] starts cluster in parallel for example.
> >>>>>>>
> >>>>>>> [1]
> >>>>>>
> >>>
> https://github.com/apache/ignite/pull/7967/files#diff-59adde2a2ab7dc17aea6c65153dfcda7R84
> >>>
> >>>>>>> [2]
> >>>>>>
> >>>
> https://github.com/apache/ignite/pull/7967/files#diff-d6a7b19f30f349d426b8894a40389cf5R79
> >>>
> >>>>>>>
> >>>>>>> On Thu, Jul 2, 2020 at 1:00 PM Nikolay Izhikov <
> [hidden email]>
> >>>>>> wrote:
> >>>>>>> Hello, Maxim.
> >>>>>>>
> >>>>>>>> 1. tiden can deploy artifacts by itself, while ducktape relies on
> >>>>>> dependencies being deployed by external scripts
> >>>>>>>
> >>>>>>> Why do you think that maintaining deploy scripts coupled with the
> >>>>>> testing framework is an advantage?
> >>>>>>> I thought we want to see and maintain deployment scripts separate
> >>>>>>> from
> >>>>>> the testing framework.
> >>>>>>>
> >>>>>>>> 2. tiden can execute actions over remote nodes in real parallel
> >>>>>> fashion, while ducktape internally does all actions sequentially.
> >>>>>>>
> >>>>>>> Can you, please, clarify, what actions do you have in mind?
> >>>>>>> And why we want to execute them concurrently?
> >>>>>>> Ignite node start, Client application execution can be done
> >>> concurrently
> >>>>>> with the ducktape approach.
> >>>>>>>
> >>>>>>>> If we used ducktape solution we would have to instead prepare some
> >>>>>> deployment scripts to pre-initialize Sberbank hosts, for example,
> >>>>>> with
> >>>>>> Ansible or Chef
> >>>>>>>
> >>>>>>> We shouldn’t take some user approach as an argument in this
> >>> discussion.
> >>>>>> Let’s discuss a general approach for all users of the Ignite.
> Anyway,
> >>> what
> >>>>>> is wrong with the external deployment script approach?
> >>>>>>>
> >>>>>>> We, as a community, should provide several ways to run integration
> >>> tests
> >>>>>> out-of-the-box AND the ability to customize deployment regarding the
> >>> user
> >>>>>> landscape.
> >>>>>>>
> >>>>>>>> You have solved this deficiency with docker by putting all
> >>>>>> dependencies into one uber-image and that looks like simple and
> >>>>>> elegant
> >>>>>> solution however, that effectively limits you to single-host
> testing.
> >>>>>>>
> >>>>>>> Docker image should be used only by the Ignite developers to test
> >>>>>> something locally.
> >>>>>>> It’s not intended for some real-world testing.
> >>>>>>>
> >>>>>>> The main issue with the Tiden that I see, it tested and
> >>>>>>> maintained as
> >>> a
> >>>>>> closed source solution.
> >>>>>>> This can lead to the hard to solve problems when we start using and
> >>>>>> maintaining it as an open-source solution.
> >>>>>>> Like, how many developers used Tiden? And how many of developers
> >>>>>>> were
> >>>>>> not authors of the Tiden itself?
> >>>>>>>
> >>>>>>>
> >>>>>>>> 2 июля 2020 г., в 12:30, Max Shonichev <[hidden email]>
> >>>>>> написал(а):
> >>>>>>>>
> >>>>>>>> Anton, Nikolay,
> >>>>>>>>
> >>>>>>>> Let's agree on what we are arguing about: whether it is about
> "like
> >>> or
> >>>>>> don't like" or about technical properties of suggested solutions.
> >>>>>>>>
> >>>>>>>> If it is about likes and dislikes, then the whole discussion is
> >>>>>> meaningless. However, I hope together we can analyse pros and cons
> >>>>>> carefully.
> >>>>>>>>
> >>>>>>>> As far as I can understand now, two main differences between
> >>>>>>>> ducktape
> >>>>>> and tiden is that:
> >>>>>>>>
> >>>>>>>> 1. tiden can deploy artifacts by itself, while ducktape relies on
> >>>>>> dependencies being deployed by external scripts.
> >>>>>>>>
> >>>>>>>> 2. tiden can execute actions over remote nodes in real parallel
> >>>>>> fashion, while ducktape internally does all actions sequentially.
> >>>>>>>>
> >>>>>>>> As for me, these are very important properties for distributed
> >>> testing
> >>>>>> framework.
> >>>>>>>>
> >>>>>>>> First property let us easily reuse tiden in existing
> >>>>>>>> infrastructures,
> >>>>>> for example, during Zookeeper IEP testing at Sberbank site we used
> >>>>>> the
> >>> same
> >>>>>> tiden scripts that we use in our lab, the only change was putting a
> >>> list of
> >>>>>> hosts into config.
> >>>>>>>>
> >>>>>>>> If we used ducktape solution we would have to instead prepare some
> >>>>>> deployment scripts to pre-initialize Sberbank hosts, for example,
> >>>>>> with
> >>>>>> Ansible or Chef.
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> You have solved this deficiency with docker by putting all
> >>>>>> dependencies into one uber-image and that looks like simple and
> >>>>>> elegant
> >>>>>> solution,
> >>>>>>>> however, that effectively limits you to single-host testing.
> >>>>>>>>
> >>>>>>>> I guess we all know about docker hyped ability to run over
> >>> distributed
> >>>>>> virtual networks. We used to go that way, but quickly found that
> >>>>>> it is
> >>> more
> >>>>>> of the hype than real work. In real environments, there are problems
> >>> with
> >>>>>> routing, DNS, multicast and broadcast traffic, and many others, that
> >>> turn
> >>>>>> docker-based distributed solution into a fragile hard-to-maintain
> >>> monster.
> >>>>>>>>
> >>>>>>>> Please, if you believe otherwise, perform a run of your PoC over
> at
> >>>>>> least two physical hosts and share results with us.
> >>>>>>>>
> >>>>>>>> If you consider that one physical docker host is enough, please,
> >>> don't
> >>>>>> overlook that we want to run real scale scenarios, with 50-100 cache
> >>>>>> groups, persistence enabled and a millions of keys loaded.
> >>>>>>>>
> >>>>>>>> Practical limit for such configurations is 4-6 nodes per single
> >>>>>> physical host. Otherwise, tests become flaky due to resource
> >>> starvation.
> >>>>>>>>
> >>>>>>>> Please, if you believe otherwise, perform at least a 10 of runs of
> >>>>>> your PoC with other tests running at TC (we're targeting TeamCity,
> >>> right?)
> >>>>>> and share results so we could check if the numbers are reproducible.
> >>>>>>>>
> >>>>>>>> I stress this once more: functional integration tests are OK to
> run
> >>> in
> >>>>>> Docker and CI, but running benchmarks in Docker is a big NO GO.
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> Second property let us write tests that require real-parallel
> >>>>>>>> actions
> >>>>>> over hosts.
> >>>>>>>>
> >>>>>>>> For example, agreed scenario for PME benchmarkduring "PME
> >>> optimization
> >>>>>> stream" was as follows:
> >>>>>>>>
> >>>>>>>>   - 10 server nodes, preloaded with 1M of keys
> >>>>>>>>   - 4 client nodes perform transactional load  (client nodes
> >>> physically
> >>>>>> separated from server nodes)
> >>>>>>>>   - during load:
> >>>>>>>>   -- 5 server nodes stopped in parallel
> >>>>>>>>   -- after 1 minute, all 5 nodes are started in parallel
> >>>>>>>>   - load stopped, logs are analysed for exchange times.
> >>>>>>>>
> >>>>>>>> If we had stopped and started 5 nodes one-by-one, as ducktape
> does,
> >>>>>> then partition map exchange merge would not happen and we could not
> >>> have
> >>>>>> measured PME optimizations for that case.
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> These are limitations of ducktape that we believe as a more
> >>>>>>>> important
> >>>>>>>> argument "against" than you provide "for".
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> On 30.06.2020 14:58, Anton Vinogradov wrote:
> >>>>>>>>> Folks,
> >>>>>>>>> First, I've created PR [1] with ducktests improvements
> >>>>>>>>> PR contains the following changes
> >>>>>>>>> - Pme-free switch proof-benchmark (2.7.6 vs master)
> >>>>>>>>> - Ability to check (compare with) previous releases (eg. 2.7.6 &
> >>> 2.8)
> >>>>>>>>> - Global refactoring
> >>>>>>>>> -- benchmarks javacode simplification
> >>>>>>>>> -- services python and java classes code deduplication
> >>>>>>>>> -- fail-fast checks for java and python (eg. application should
> >>>>>> explicitly write it finished with success)
> >>>>>>>>> -- simple results extraction from tests and benchmarks
> >>>>>>>>> -- javacode now configurable from tests/benchmarks
> >>>>>>>>> -- proper SIGTERM handling at javacode (eg. it may finish last
> >>>>>> operation and log results)
> >>>>>>>>> -- docker volume now marked as delegated to increase execution
> >>>>>>>>> speed
> >>>>>> for mac & win users
> >>>>>>>>> -- Ignite cluster now start in parallel (start speed-up)
> >>>>>>>>> -- Ignite can be configured at test/benchmark
> >>>>>>>>> - full and module assembly scripts added
> >>>>>>>> Great job done! But let me remind one of Apache Ignite principles:
> >>>>>>>> week of thinking save months of development.
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>> Second, I'd like to propose to accept ducktests [2] (ducktape
> >>>>>> integration) as a target "PoC check & real topology benchmarking
> >>>>>> tool".
> >>>>>>>>> Ducktape pros
> >>>>>>>>> - Developed for distributed system by distributed system
> >>>>>>>>> developers.
> >>>>>>>> So does Tiden
> >>>>>>>>
> >>>>>>>>> - Developed since 2014, stable.
> >>>>>>>> Tiden is also pretty stable, and development start date is not a
> >>>>>>>> good
> >>>>>> argument, for example pytest is since 2004, pytest-xdist (plugin for
> >>>>>> distributed testing) is since 2010, but we don't see it as a
> >>> alternative at
> >>>>>> all.
> >>>>>>>>
> >>>>>>>>> - Proven usability by usage at Kafka.
> >>>>>>>> Tiden is proven usable by usage at GridGain and Sberbank
> >>>>>>>> deployments.
> >>>>>>>> Core, storage, sql and tx teams use benchmark results provided by
> >>>>>> Tiden on a daily basis.
> >>>>>>>>
> >>>>>>>>> - Dozens of dozens tests and benchmarks at Kafka as a great
> >>>>>>>>> example
> >>>>>> pack.
> >>>>>>>> We'll donate some of our suites to Ignite as I've mentioned in
> >>>>>> previous letter.
> >>>>>>>>
> >>>>>>>>> - Built-in Docker support for rapid development and checks.
> >>>>>>>> False, there's no specific 'docker support' in ducktape itself,
> you
> >>>>>> just wrap it in docker by yourself, because ducktape is lacking
> >>> deployment
> >>>>>> abilities.
> >>>>>>>>
> >>>>>>>>> - Great for CI automation.
> >>>>>>>> False, there's no specific CI-enabled features in ducktape.
> >>>>>>>> Tiden, on
> >>>>>> the other hand, provide generic xUnit reporting format, which is
> >>> supported
> >>>>>> by both TeamCity and Jenkins. Also, instead of using private keys,
> >>> Tiden
> >>>>>> can use SSH agent, which is also great for CI, because both
> >>>>>>>> TeamCity and Jenkins store keys in secret storage available only
> >>>>>>>> for
> >>>>>> ssh-agent and only for the time of the test.
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>>> As an additional motivation, at least 3 teams
> >>>>>>>>> - IEP-45 team (to check crash-recovery speed-up (discovery and
> >>> Zabbix
> >>>>>> speed-up))
> >>>>>>>>> - Ignite SE Plugins team (to check plugin's features does not
> >>>>>> slow-down or broke AI features)
> >>>>>>>>> - Ignite SE QA team (to append already developed
> >>>>>>>>> smoke/load/failover
> >>>>>> tests to AI codebase)
> >>>>>>>>
> >>>>>>>> Please, before recommending your tests to other teams, provide
> >>>>>>>> proofs
> >>>>>>>> that your tests are reproducible in real environment.
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>> now, wait for ducktest merge to start checking cases they
> >>>>>>>>> working on
> >>>>>> in AI way.
> >>>>>>>>> Thoughts?
> >>>>>>>> Let us together review both solutions, we'll try to run your
> >>>>>>>> tests in
> >>>>>> our lab, and you'll try to at least checkout tiden and see if same
> >>> tests
> >>>>>> can be implemented with it?
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>> [1] https://github.com/apache/ignite/pull/7967
> >>>>>>>>> [2] https://github.com/apache/ignite/tree/ignite-ducktape
> >>>>>>>>> On Tue, Jun 16, 2020 at 12:22 PM Nikolay Izhikov <
> >>> [hidden email]
> >>>>>> <mailto:[hidden email]>> wrote:
> >>>>>>>>>     Hello, Maxim.
> >>>>>>>>>     Thank you for so detailed explanation.
> >>>>>>>>>     Can we put the content of this discussion somewhere on the
> >>>>>>>>> wiki?
> >>>>>>>>>     So It doesn’t get lost.
> >>>>>>>>>     I divide the answer in several parts. From the requirements
> to
> >>> the
> >>>>>>>>>     implementation.
> >>>>>>>>>     So, if we agreed on the requirements we can proceed with the
> >>>>>>>>>     discussion of the implementation.
> >>>>>>>>>     1. Requirements:
> >>>>>>>>>     The main goal I want to achieve is *reproducibility* of the
> >>> tests.
> >>>>>>>>>     I’m sick and tired with the zillions of flaky, rarely
> >>>>>>>>> failed, and
> >>>>>>>>>     almost never failed tests in Ignite codebase.
> >>>>>>>>>     We should start with the simplest scenarios that will be as
> >>>>>> reliable
> >>>>>>>>>     as steel :)
> >>>>>>>>>     I want to know for sure:
> >>>>>>>>>        - Is this PR makes rebalance quicker or not?
> >>>>>>>>>        - Is this PR makes PME quicker or not?
> >>>>>>>>>     So, your description of the complex test scenario looks as
> >>>>>>>>> a next
> >>>>>>>>>     step to me.
> >>>>>>>>>     Anyway, It’s cool we already have one.
> >>>>>>>>>     The second goal is to have a strict test lifecycle as we
> >>>>>>>>> have in
> >>>>>>>>>     JUnit and similar frameworks.
> >>>>>>>>>      > It covers production-like deployment and running a
> >>>>>>>>> scenarios
> >>>>>> over
> >>>>>>>>>     a single database instance.
> >>>>>>>>>     Do you mean «single cluster» or «single host»?
> >>>>>>>>>     2. Existing tests:
> >>>>>>>>>      > A Combinator suite allows to run set of operations
> >>> concurrently
> >>>>>>>>>     over given database instance.
> >>>>>>>>>      > A Consumption suite allows to run a set production-like
> >>> actions
> >>>>>>>>>     over given set of Ignite/GridGain versions and compare test
> >>> metrics
> >>>>>>>>>     across versions
> >>>>>>>>>      > A Yardstick suite
> >>>>>>>>>      > A Stress suite that simulates hardware environment
> >>>>>>>>> degradation
> >>>>>>>>>      > An Ultimate, DR and Compatibility suites that performs
> >>>>>> functional
> >>>>>>>>>     regression testing
> >>>>>>>>>      > Regression
> >>>>>>>>>     Great news that we already have so many choices for testing!
> >>>>>>>>>     Mature test base is a big +1 for Tiden.
> >>>>>>>>>     3. Comparison:
> >>>>>>>>>      > Criteria: Test configuration
> >>>>>>>>>      > Ducktape: single JSON string for all tests
> >>>>>>>>>      > Tiden: any number of YaML config files, command line
> option
> >>> for
> >>>>>>>>>     fine-grained test configuration, ability to select/modify
> >>>>>>>>> tests
> >>>>>>>>>     behavior based on Ignite version.
> >>>>>>>>>     1. Many YAML files can be hard to maintain.
> >>>>>>>>>     2. In ducktape, you can set parameters via «—parameters»
> >>>>>>>>> option.
> >>>>>>>>>     Please, take a look at the doc [1]
> >>>>>>>>>      > Criteria: Cluster control
> >>>>>>>>>      > Tiden: additionally can address cluster as a whole and
> >>>>>>>>> execute
> >>>>>>>>>     remote commands in parallel.
> >>>>>>>>>     It seems we implement this ability in the PoC, already.
> >>>>>>>>>      > Criteria: Test assertions
> >>>>>>>>>      > Tiden: simple asserts, also few customized assertion
> >>>>>>>>> helpers.
> >>>>>>>>>      > Ducktape: simple asserts.
> >>>>>>>>>     Can you, please, be more specific.
> >>>>>>>>>     What helpers do you have in mind?
> >>>>>>>>>     Ducktape has an asserts that waits for logfile messages or
> >>>>>>>>> some
> >>>>>>>>>     process finish.
> >>>>>>>>>      > Criteria: Test reporting
> >>>>>>>>>      > Ducktape: limited to its own text/HTML format
> >>>>>>>>>     Ducktape have
> >>>>>>>>>     1. Text reporter
> >>>>>>>>>     2. Customizable HTML reporter
> >>>>>>>>>     3. JSON reporter.
> >>>>>>>>>     We can show JSON with the any template or tool.
> >>>>>>>>>      > Criteria: Provisioning and deployment
> >>>>>>>>>      > Ducktape: can provision subset of hosts from cluster for
> >>>>>>>>> test
> >>>>>>>>>     needs. However, that means, that test can’t be scaled without
> >>> test
> >>>>>>>>>     code changes. Does not do any deploy, relies on external
> >>>>>>>>> means,
> >>>>>> e.g.
> >>>>>>>>>     pre-packaged in docker image, as in PoC.
> >>>>>>>>>     This is not true.
> >>>>>>>>>     1. We can set explicit test parameters(node number) via
> >>> parameters.
> >>>>>>>>>     We can increase client count of cluster size without test
> code
> >>>>>> changes.
> >>>>>>>>>     2. We have many choices for the test environment. These
> >>>>>>>>> choices
> >>> are
> >>>>>>>>>     tested and used in other projects:
> >>>>>>>>>              * docker
> >>>>>>>>>              * vagrant
> >>>>>>>>>              * private cloud(ssh access)
> >>>>>>>>>              * ec2
> >>>>>>>>>     Please, take a look at Kafka documentation [2]
> >>>>>>>>>      > I can continue more on this, but it should be enough for
> >>>>>>>>> now:
> >>>>>>>>>     We need to go deeper! :)
> >>>>>>>>>     [1]
> >>>>>>>>>
> >>>>>>
> https://ducktape-docs.readthedocs.io/en/latest/run_tests.html#options
> >>>>>>>>>     [2]
> >>>>>> https://github.com/apache/kafka/tree/trunk/tests#ec2-quickstart
> >>>>>>>>>      > 9 июня 2020 г., в 17:25, Max A. Shonichev
> >>>>>>>>> <[hidden email]
> >>>>>>>>>     <mailto:[hidden email]>> написал(а):
> >>>>>>>>>      >
> >>>>>>>>>      > Greetings, Nikolay,
> >>>>>>>>>      >
> >>>>>>>>>      > First of all, thank you for you great effort preparing
> >>>>>>>>> PoC of
> >>>>>>>>>     integration testing to Ignite community.
> >>>>>>>>>      >
> >>>>>>>>>      > It’s a shame Ignite did not have at least some such
> >>>>>>>>> tests yet,
> >>>>>>>>>     however, GridGain, as a major contributor to Apache Ignite
> >>>>>>>>> had a
> >>>>>>>>>     profound collection of in-house tools to perform
> >>>>>>>>> integration and
> >>>>>>>>>     performance testing for years already and while we slowly
> >>> consider
> >>>>>>>>>     sharing our expertise with the community, your initiative
> >>>>>>>>> makes
> >>> us
> >>>>>>>>>     drive that process a bit faster, thanks a lot!
> >>>>>>>>>      >
> >>>>>>>>>      > I reviewed your PoC and want to share a little about
> >>>>>>>>> what we
> >>> do
> >>>>>>>>>     on our part, why and how, hope it would help community take
> >>> proper
> >>>>>>>>>     course.
> >>>>>>>>>      >
> >>>>>>>>>      > First I’ll do a brief overview of what decisions we made
> >>>>>>>>> and
> >>>>>> what
> >>>>>>>>>     we do have in our private code base, next I’ll describe
> >>>>>>>>> what we
> >>>>>> have
> >>>>>>>>>     already donated to the public and what we plan public next,
> >>>>>>>>> then
> >>>>>>>>>     I’ll compare both approaches highlighting deficiencies in
> >>>>>>>>> order
> >>> to
> >>>>>>>>>     spur public discussion on the matter.
> >>>>>>>>>      >
> >>>>>>>>>      > It might seem strange to use Python to run Bash to run
> Java
> >>>>>>>>>     applications because that introduces IT industry best of
> >>>>>>>>> breed’ –
> >>>>>>>>>     the Python dependency hell – to the Java application code
> >>>>>>>>> base.
> >>> The
> >>>>>>>>>     only strangest decision one can made is to use Maven to run
> >>> Docker
> >>>>>>>>>     to run Bash to run Python to run Bash to run Java, but
> >>>>>>>>> desperate
> >>>>>>>>>     times call for desperate measures I guess.
> >>>>>>>>>      >
> >>>>>>>>>      > There are Java-based solutions for integration testing
> >>>>>>>>> exists,
> >>>>>>>>>     e.g. Testcontainers [1], Arquillian [2], etc, and they
> >>>>>>>>> might go
> >>>>>> well
> >>>>>>>>>     for Ignite community CI pipelines by them selves. But we also
> >>>>>> wanted
> >>>>>>>>>     to run performance tests and benchmarks, like the dreaded PME
> >>>>>>>>>     benchmark, and this is solved by totally different set of
> >>>>>>>>> tools
> >>> in
> >>>>>>>>>     Java world, e.g. Jmeter [3], OpenJMH [4], Gatling [5], etc.
> >>>>>>>>>      >
> >>>>>>>>>      > Speaking specifically about benchmarking, Apache Ignite
> >>>>>> community
> >>>>>>>>>     already has Yardstick [6], and there’s nothing wrong with
> >>>>>>>>> writing
> >>>>>>>>>     PME benchmark using Yardstick, but we also wanted to be
> >>>>>>>>> able to
> >>> run
> >>>>>>>>>     scenarios like this:
> >>>>>>>>>      > - put an X load to a Ignite database;
> >>>>>>>>>      > - perform an Y set of operations to check how Ignite copes
> >>> with
> >>>>>>>>>     operations under load.
> >>>>>>>>>      >
> >>>>>>>>>      > And yes, we also wanted applications under test be
> deployed
> >>>>>> ‘like
> >>>>>>>>>     in a production’, e.g. distributed over a set of hosts. This
> >>> arises
> >>>>>>>>>     questions about provisioning and nodes affinity which I’ll
> >>>>>>>>> cover
> >>> in
> >>>>>>>>>     detail later.
> >>>>>>>>>      >
> >>>>>>>>>      > So we decided to put a little effort to build a simple
> >>>>>>>>> tool to
> >>>>>>>>>     cover different integration and performance scenarios, and
> >>>>>>>>> our QA
> >>>>>>>>>     lab first attempt was PoC-Tester [7], currently open source
> >>>>>>>>> for
> >>> all
> >>>>>>>>>     but for reporting web UI. It’s a quite simple to use 95%
> >>> Java-based
> >>>>>>>>>     tool targeted to be run on a pre-release QA stage.
> >>>>>>>>>      >
> >>>>>>>>>      > It covers production-like deployment and running a
> >>>>>>>>> scenarios
> >>>>>> over
> >>>>>>>>>     a single database instance. PoC-Tester scenarios consists of
> a
> >>>>>>>>>     sequence of tasks running sequentially or in parallel.
> >>>>>>>>> After all
> >>>>>>>>>     tasks complete, or at any time during test, user can run logs
> >>>>>>>>>     collection task, logs are checked against exceptions and a
> >>> summary
> >>>>>>>>>     of found issues and task ops/latency statistics is
> >>>>>>>>> generated at
> >>> the
> >>>>>>>>>     end of scenario. One of the main PoC-Tester features is its
> >>>>>>>>>     fire-and-forget approach to task managing. That is, you can
> >>> deploy
> >>>>>> a
> >>>>>>>>>     grid and left it running for weeks, periodically firing some
> >>> tasks
> >>>>>>>>>     onto it.
> >>>>>>>>>      >
> >>>>>>>>>      > During earliest stages of PoC-Tester development it
> becomes
> >>>>>> quite
> >>>>>>>>>     clear that Java application development is a tedious
> >>>>>>>>> process and
> >>>>>>>>>     architecture decisions you take during development are slow
> >>>>>>>>> and
> >>>>>> hard
> >>>>>>>>>     to change.
> >>>>>>>>>      > For example, scenarios like this
> >>>>>>>>>      > - deploy two instances of GridGain with master-slave data
> >>>>>>>>>     replication configured;
> >>>>>>>>>      > - put a load on master;
> >>>>>>>>>      > - perform checks on slave,
> >>>>>>>>>      > or like this:
> >>>>>>>>>      > - preload a 1Tb of data by using your favorite tool of
> >>>>>>>>> choice
> >>> to
> >>>>>>>>>     an Apache Ignite of version X;
> >>>>>>>>>      > - run a set of functional tests running Apache Ignite
> >>>>>>>>> version
> >>> Y
> >>>>>>>>>     over preloaded data,
> >>>>>>>>>      > do not fit well in the PoC-Tester workflow.
> >>>>>>>>>      >
> >>>>>>>>>      > So, this is why we decided to use Python as a generic
> >>> scripting
> >>>>>>>>>     language of choice.
> >>>>>>>>>      >
> >>>>>>>>>      > Pros:
> >>>>>>>>>      > - quicker prototyping and development cycles
> >>>>>>>>>      > - easier to find DevOps/QA engineer with Python skills
> than
> >>> one
> >>>>>>>>>     with Java skills
> >>>>>>>>>      > - used extensively all over the world for DevOps/CI
> >>>>>>>>> pipelines
> >>>>>> and
> >>>>>>>>>     thus has rich set of libraries for all possible integration
> >>>>>>>>> uses
> >>>>>> cases.
> >>>>>>>>>      >
> >>>>>>>>>      > Cons:
> >>>>>>>>>      > - Nightmare with dependencies. Better stick to specific
> >>>>>>>>>     language/libraries version.
> >>>>>>>>>      >
> >>>>>>>>>      > Comparing alternatives for Python-based testing
> >>>>>>>>> framework we
> >>>>>> have
> >>>>>>>>>     considered following requirements, somewhat similar to what
> >>> you’ve
> >>>>>>>>>     mentioned for Confluent [8] previously:
> >>>>>>>>>      > - should be able run locally or distributed (bare metal
> >>>>>>>>> or in
> >>>>>> the
> >>>>>>>>>     cloud)
> >>>>>>>>>      > - should have built-in deployment facilities for
> >>>>>>>>> applications
> >>>>>>>>>     under test
> >>>>>>>>>      > - should separate test configuration and test code
> >>>>>>>>>      > -- be able to easily reconfigure tests by simple
> >>>>>>>>> configuration
> >>>>>>>>>     changes
> >>>>>>>>>      > -- be able to easily scale test environment by simple
> >>>>>>>>>     configuration changes
> >>>>>>>>>      > -- be able to perform regression testing by simple
> >>>>>>>>> switching
> >>>>>>>>>     artifacts under test via configuration
> >>>>>>>>>      > -- be able to run tests with different JDK version by
> >>>>>>>>> simple
> >>>>>>>>>     configuration changes
> >>>>>>>>>      > - should have human readable reports and/or reporting
> tools
> >>>>>>>>>     integration
> >>>>>>>>>      > - should allow simple test progress monitoring, one does
> >>>>>>>>> not
> >>>>>> want
> >>>>>>>>>     to run 6-hours test to find out that application actually
> >>>>>>>>> crashed
> >>>>>>>>>     during first hour.
> >>>>>>>>>      > - should allow parallel execution of test actions
> >>>>>>>>>      > - should have clean API for test writers
> >>>>>>>>>      > -- clean API for distributed remote commands execution
> >>>>>>>>>      > -- clean API for deployed applications start / stop and
> >>>>>>>>> other
> >>>>>>>>>     operations
> >>>>>>>>>      > -- clean API for performing check on results
> >>>>>>>>>      > - should be open source or at least source code should
> >>>>>>>>> allow
> >>>>>> ease
> >>>>>>>>>     change or extension
> >>>>>>>>>      >
> >>>>>>>>>      > Back at that time we found no better alternative than to
> >>>>>>>>> write
> >>>>>>>>>     our own framework, and here goes Tiden [9] as GridGain
> >>>>>>>>> framework
> >>> of
> >>>>>>>>>     choice for functional integration and performance testing.
> >>>>>>>>>      >
> >>>>>>>>>      > Pros:
> >>>>>>>>>      > - solves all the requirements above
> >>>>>>>>>      > Cons (for Ignite):
> >>>>>>>>>      > - (currently) closed GridGain source
> >>>>>>>>>      >
> >>>>>>>>>      > On top of Tiden we’ve built a set of test suites, some of
> >>> which
> >>>>>>>>>     you might have heard already.
> >>>>>>>>>      >
> >>>>>>>>>      > A Combinator suite allows to run set of operations
> >>> concurrently
> >>>>>>>>>     over given database instance. Proven to find at least 30+
> race
> >>>>>>>>>     conditions and NPE issues.
> >>>>>>>>>      >
> >>>>>>>>>      > A Consumption suite allows to run a set production-like
> >>> actions
> >>>>>>>>>     over given set of Ignite/GridGain versions and compare test
> >>> metrics
> >>>>>>>>>     across versions, like heap/disk/CPU consumption, time to
> >>>>>>>>> perform
> >>>>>>>>>     actions, like client PME, server PME, rebalancing time, data
> >>>>>>>>>     replication time, etc.
> >>>>>>>>>      >
> >>>>>>>>>      > A Yardstick suite is a thin layer of Python glue code to
> >>>>>>>>> run
> >>>>>>>>>     Apache Ignite pre-release benchmarks set. Yardstick itself
> >>>>>>>>> has a
> >>>>>>>>>     mediocre deployment capabilities, Tiden solves this easily.
> >>>>>>>>>      >
> >>>>>>>>>      > A Stress suite that simulates hardware environment
> >>>>>>>>> degradation
> >>>>>>>>>     during testing.
> >>>>>>>>>      >
> >>>>>>>>>      > An Ultimate, DR and Compatibility suites that performs
> >>>>>> functional
> >>>>>>>>>     regression testing of GridGain Ultimate Edition features like
> >>>>>>>>>     snapshots, security, data replication, rolling upgrades, etc.
> >>>>>>>>>      >
> >>>>>>>>>      > A Regression and some IEPs testing suites, like IEP-14,
> >>> IEP-15,
> >>>>>>>>>     etc, etc, etc.
> >>>>>>>>>      >
> >>>>>>>>>      > Most of the suites above use another in-house developed
> >>>>>>>>> Java
> >>>>>> tool
> >>>>>>>>>     – PiClient – to perform actual loading and miscellaneous
> >>> operations
> >>>>>>>>>     with Ignite under test. We use py4j Python-Java gateway
> >>>>>>>>> library
> >>> to
> >>>>>>>>>     control PiClient instances from the tests.
> >>>>>>>>>      >
> >>>>>>>>>      > When we considered CI, we put TeamCity out of scope,
> >>>>>>>>> because
> >>>>>>>>>     distributed integration and performance tests tend to run for
> >>> hours
> >>>>>>>>>     and TeamCity agents are scarce and costly resource. So,
> >>>>>>>>> bundled
> >>>>>> with
> >>>>>>>>>     Tiden there is jenkins-job-builder [10] based CI pipelines
> and
> >>>>>>>>>     Jenkins xUnit reporting. Also, rich web UI tool Ward
> >>>>>>>>> aggregates
> >>>>>> test
> >>>>>>>>>     run reports across versions and has built in visualization
> >>> support
> >>>>>>>>>     for Combinator suite.
> >>>>>>>>>      >
> >>>>>>>>>      > All of the above is currently closed source, but we plan
> to
> >>> make
> >>>>>>>>>     it public for community, and publishing Tiden core [9] is the
> >>> first
> >>>>>>>>>     step on that way. You can review some examples of using
> >>>>>>>>> Tiden for
> >>>>>>>>>     tests at my repository [11], for start.
> >>>>>>>>>      >
> >>>>>>>>>      > Now, let’s compare Ducktape PoC and Tiden.
> >>>>>>>>>      >
> >>>>>>>>>      > Criteria: Language
> >>>>>>>>>      > Tiden: Python, 3.7
> >>>>>>>>>      > Ducktape: Python, proposes itself as Python 2.7, 3.6, 3.7
> >>>>>>>>>     compatible, but actually can’t work with Python 3.7 due to
> >>>>>>>>> broken
> >>>>>>>>>     Zmq dependency.
> >>>>>>>>>      > Comment: Python 3.7 has a much better support for
> >>>>>>>>> async-style
> >>>>>>>>>     code which might be crucial for distributed application
> >>>>>>>>> testing.
> >>>>>>>>>      > Score: Tiden: 1, Ducktape: 0
> >>>>>>>>>      >
> >>>>>>>>>      > Criteria: Test writers API
> >>>>>>>>>      > Supported integration test framework concepts are
> basically
> >>> the
> >>>>>> same:
> >>>>>>>>>      > - a test controller (test runner)
> >>>>>>>>>      > - a cluster
> >>>>>>>>>      > - a node
> >>>>>>>>>      > - an application (a service in Ducktape terms)
> >>>>>>>>>      > - a test
> >>>>>>>>>      > Score: Tiden: 5, Ducktape: 5
> >>>>>>>>>      >
> >>>>>>>>>      > Criteria: Tests selection and run
> >>>>>>>>>      > Ducktape: suite-package-class-method level selection,
> >>>>>>>>> internal
> >>>>>>>>>     scheduler allows to run tests in suite in parallel.
> >>>>>>>>>      > Tiden: also suite-package-class-method level selection,
> >>>>>>>>>     additionally allows selecting subset of tests by attribute,
> >>>>>> parallel
> >>>>>>>>>     runs not built in, but allows merging test reports after
> >>> different
> >>>>>> runs.
> >>>>>>>>>      > Score: Tiden: 2, Ducktape: 2
> >>>>>>>>>      >
> >>>>>>>>>      > Criteria: Test configuration
> >>>>>>>>>      > Ducktape: single JSON string for all tests
> >>>>>>>>>      > Tiden: any number of YaML config files, command line
> option
> >>> for
> >>>>>>>>>     fine-grained test configuration, ability to select/modify
> >>>>>>>>> tests
> >>>>>>>>>     behavior based on Ignite version.
> >>>>>>>>>      > Score: Tiden: 3, Ducktape: 1
> >>>>>>>>>      >
> >>>>>>>>>      > Criteria: Cluster control
> >>>>>>>>>      > Ducktape: allow execute remote commands by node
> granularity
> >>>>>>>>>      > Tiden: additionally can address cluster as a whole and
> >>>>>>>>> execute
> >>>>>>>>>     remote commands in parallel.
> >>>>>>>>>      > Score: Tiden: 2, Ducktape: 1
> >>>>>>>>>      >
> >>>>>>>>>      > Criteria: Logs control
> >>>>>>>>>      > Both frameworks have similar builtin support for remote
> >>>>>>>>> logs
> >>>>>>>>>     collection and grepping. Tiden has built-in plugin that can
> >>>>>>>>> zip,
> >>>>>>>>>     collect arbitrary log files from arbitrary locations at
> >>>>>>>>>     test/module/suite granularity and unzip if needed, also
> >>> application
> >>>>>>>>>     API to search / wait for messages in logs. Ducktape allows
> >>>>>>>>> each
> >>>>>>>>>     service declare its log files location (seemingly does not
> >>> support
> >>>>>>>>>     logs rollback), and a single entrypoint to collect service
> >>>>>>>>> logs.
> >>>>>>>>>      > Score: Tiden: 1, Ducktape: 1
> >>>>>>>>>      >
> >>>>>>>>>      > Criteria: Test assertions
> >>>>>>>>>      > Tiden: simple asserts, also few customized assertion
> >>>>>>>>> helpers.
> >>>>>>>>>      > Ducktape: simple asserts.
> >>>>>>>>>      > Score: Tiden: 2, Ducktape: 1
> >>>>>>>>>      >
> >>>>>>>>>      > Criteria: Test reporting
> >>>>>>>>>      > Ducktape: limited to its own text/html format
> >>>>>>>>>      > Tiden: provides text report, yaml report for reporting
> >>>>>>>>> tools
> >>>>>>>>>     integration, XML xUnit report for integration with
> >>>>>> Jenkins/TeamCity.
> >>>>>>>>>      > Score: Tiden: 3, Ducktape: 1
> >>>>>>>>>      >
> >>>>>>>>>      > Criteria: Provisioning and deployment
> >>>>>>>>>      > Ducktape: can provision subset of hosts from cluster for
> >>>>>>>>> test
> >>>>>>>>>     needs. However, that means, that test can’t be scaled without
> >>> test
> >>>>>>>>>     code changes. Does not do any deploy, relies on external
> >>>>>>>>> means,
> >>>>>> e.g.
> >>>>>>>>>     pre-packaged in docker image, as in PoC.
> >>>>>>>>>      > Tiden: Given a set of hosts, Tiden uses all of them for
> the
> >>>>>> test.
> >>>>>>>>>     Provisioning should be done by external means. However,
> >>>>>>>>> provides
> >>> a
> >>>>>>>>>     conventional automated deployment routines.
> >>>>>>>>>      > Score: Tiden: 1, Ducktape: 1
> >>>>>>>>>      >
> >>>>>>>>>      > Criteria: Documentation and Extensibility
> >>>>>>>>>      > Tiden: current API documentation is limited, should
> >>>>>>>>> change as
> >>> we
> >>>>>>>>>     go open source. Tiden is easily extensible via hooks and
> >>>>>>>>> plugins,
> >>>>>>>>>     see example Maven plugin and Gatling application at [11].
> >>>>>>>>>      > Ducktape: basic documentation at readthedocs.io
> >>>>>>>>>     <http://readthedocs.io>. Codebase is rigid, framework core
> is
> >>>>>>>>>     tightly coupled and hard to change. The only possible
> >>>>>>>>> extension
> >>>>>>>>>     mechanism is fork-and-rewrite.
> >>>>>>>>>      > Score: Tiden: 2, Ducktape: 1
> >>>>>>>>>      >
> >>>>>>>>>      > I can continue more on this, but it should be enough for
> >>>>>>>>> now:
> >>>>>>>>>      > Overall score: Tiden: 22, Ducktape: 14.
> >>>>>>>>>      >
> >>>>>>>>>      > Time for discussion!
> >>>>>>>>>      >
> >>>>>>>>>      > ---
> >>>>>>>>>      > [1] - https://www.testcontainers.org/
> >>>>>>>>>      > [2] - http://arquillian.org/guides/getting_started/
> >>>>>>>>>      > [3] - https://jmeter.apache.org/index.html
> >>>>>>>>>      > [4] - https://openjdk.java.net/projects/code-tools/jmh/
> >>>>>>>>>      > [5] - https://gatling.io/docs/current/
> >>>>>>>>>      > [6] - https://github.com/gridgain/yardstick
> >>>>>>>>>      > [7] - https://github.com/gridgain/poc-tester
> >>>>>>>>>      > [8] -
> >>>>>>>>>
> >>>>>>
> >>>
> https://cwiki.apache.org/confluence/display/KAFKA/System+Test+Improvements
> >>>
> >>>>>>>>>      > [9] - https://github.com/gridgain/tiden
> >>>>>>>>>      > [10] - https://pypi.org/project/jenkins-job-builder/
> >>>>>>>>>      > [11] - https://github.com/mshonichev/tiden_examples
> >>>>>>>>>      >
> >>>>>>>>>      > On 25.05.2020 11:09, Nikolay Izhikov wrote:
> >>>>>>>>>      >> Hello,
> >>>>>>>>>      >>
> >>>>>>>>>      >> Branch with duck tape created -
> >>>>>>>>>     https://github.com/apache/ignite/tree/ignite-ducktape
> >>>>>>>>>      >>
> >>>>>>>>>      >> Any who are willing to contribute to PoC are welcome.
> >>>>>>>>>      >>
> >>>>>>>>>      >>
> >>>>>>>>>      >>> 21 мая 2020 г., в 22:33, Nikolay Izhikov
> >>>>>>>>>     <[hidden email] <mailto:[hidden email]>>
> >>>>>> написал(а):
> >>>>>>>>>      >>>
> >>>>>>>>>      >>> Hello, Denis.
> >>>>>>>>>      >>>
> >>>>>>>>>      >>> There is no rush with these improvements.
> >>>>>>>>>      >>> We can wait for Maxim proposal and compare two
> >>>>>>>>> solutions :)
> >>>>>>>>>      >>>
> >>>>>>>>>      >>>> 21 мая 2020 г., в 22:24, Denis Magda <
> [hidden email]
> >>>>>>>>>     <mailto:[hidden email]>> написал(а):
> >>>>>>>>>      >>>>
> >>>>>>>>>      >>>> Hi Nikolay,
> >>>>>>>>>      >>>>
> >>>>>>>>>      >>>> Thanks for kicking off this conversation and sharing
> >>>>>>>>> your
> >>>>>>>>>     findings with the
> >>>>>>>>>      >>>> results. That's the right initiative. I do agree that
> >>> Ignite
> >>>>>>>>>     needs to have
> >>>>>>>>>      >>>> an integration testing framework with capabilities
> >>>>>>>>> listed
> >>> by
> >>>>>> you.
> >>>>>>>>>      >>>>
> >>>>>>>>>      >>>> As we discussed privately, I would only check if
> >>>>>>>>> instead of
> >>>>>>>>>      >>>> Confluent's Ducktape library, we can use an integration
> >>>>>>>>>     testing framework
> >>>>>>>>>      >>>> developed by GridGain for testing of Ignite/GridGain
> >>>>>> clusters.
> >>>>>>>>>     That
> >>>>>>>>>      >>>> framework has been battle-tested and might be more
> >>>>>> convenient for
> >>>>>>>>>      >>>> Ignite-specific workloads. Let's wait for @Maksim
> >>>>>>>>> Shonichev
> >>>>>>>>>      >>>> <[hidden email]
> >>>>>>>>> <mailto:[hidden email]>>
> >>>>>> who
> >>>>>>>>>     promised to join this thread once he finishes
> >>>>>>>>>      >>>> preparing the usage examples of the framework. To my
> >>>>>>>>>     knowledge, Max has
> >>>>>>>>>      >>>> already been working on that for several days.
> >>>>>>>>>      >>>>
> >>>>>>>>>      >>>> -
> >>>>>>>>>      >>>> Denis
> >>>>>>>>>      >>>>
> >>>>>>>>>      >>>>
> >>>>>>>>>      >>>> On Thu, May 21, 2020 at 12:27 AM Nikolay Izhikov
> >>>>>>>>>     <[hidden email] <mailto:[hidden email]>>
> >>>>>>>>>      >>>> wrote:
> >>>>>>>>>      >>>>
> >>>>>>>>>      >>>>> Hello, Igniters.
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> I created a PoC [1] for the integration tests of
> >>>>>>>>> Ignite.
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> Let me briefly explain the gap I want to cover:
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> 1. For now, we don’t have a solution for automated
> >>>>>>>>> testing
> >>>>>> of
> >>>>>>>>>     Ignite on
> >>>>>>>>>      >>>>> «real cluster».
> >>>>>>>>>      >>>>> By «real cluster» I mean cluster «like a production»:
> >>>>>>>>>      >>>>>       * client and server nodes deployed on different
> >>> hosts.
> >>>>>>>>>      >>>>>       * thin clients perform queries from some other
> >>>>>>>>> hosts
> >>>>>>>>>      >>>>>       * etc.
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> 2. We don’t have a solution for automated benchmarks
> of
> >>> some
> >>>>>>>>>     internal
> >>>>>>>>>      >>>>> Ignite process
> >>>>>>>>>      >>>>>       * PME
> >>>>>>>>>      >>>>>       * rebalance.
> >>>>>>>>>      >>>>> This means we don’t know - Do we perform
> >>>>>>>>> rebalance(or PME)
> >>>>>> in
> >>>>>>>>>     2.7.0 faster
> >>>>>>>>>      >>>>> or slower than in 2.8.0 for the same cluster?
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> 3. We don’t have a solution for automated testing of
> >>> Ignite
> >>>>>>>>>     integration in
> >>>>>>>>>      >>>>> a real-world environment:
> >>>>>>>>>      >>>>> Ignite-Spark integration can be taken as an example.
> >>>>>>>>>      >>>>> I think some ML solutions also should be tested in
> >>>>>> real-world
> >>>>>>>>>     deployments.
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> Solution:
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> I propose to use duck tape library from confluent
> >>>>>>>>> (apache
> >>>>>> 2.0
> >>>>>>>>>     license)
> >>>>>>>>>      >>>>> I tested it both on the real cluster(Yandex Cloud)
> >>>>>>>>> and on
> >>>>>> the
> >>>>>>>>>     local
> >>>>>>>>>      >>>>> environment(docker) and it works just fine.
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> PoC contains following services:
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>>       * Simple rebalance test:
> >>>>>>>>>      >>>>>               Start 2 server nodes,
> >>>>>>>>>      >>>>>               Create some data with Ignite client,
> >>>>>>>>>      >>>>>               Start one more server node,
> >>>>>>>>>      >>>>>               Wait for rebalance finish
> >>>>>>>>>      >>>>>       * Simple Ignite-Spark integration test:
> >>>>>>>>>      >>>>>               Start 1 Spark master, start 1 Spark
> >>>>>>>>> worker,
> >>>>>>>>>      >>>>>               Start 1 Ignite server node
> >>>>>>>>>      >>>>>               Create some data with Ignite client,
> >>>>>>>>>      >>>>>               Check data in application that queries
> it
> >>> from
> >>>>>>>>>     Spark.
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> All tests are fully automated.
> >>>>>>>>>      >>>>> Logs collection works just fine.
> >>>>>>>>>      >>>>> You can see an example of the tests report - [4].
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> Pros:
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> * Ability to test local changes(no need to public
> >>>>>>>>> changes
> >>> to
> >>>>>>>>>     some remote
> >>>>>>>>>      >>>>> repository or similar).
> >>>>>>>>>      >>>>> * Ability to parametrize test environment(run the same
> >>> tests
> >>>>>>>>>     on different
> >>>>>>>>>      >>>>> JDK, JVM params, config, etc.)
> >>>>>>>>>      >>>>> * Isolation by default so system tests are as
> >>>>>>>>> reliable as
> >>>>>>>>>     possible.
> >>>>>>>>>      >>>>> * Utilities for pulling up and tearing down services
> >>> easily
> >>>>>>>>>     in clusters in
> >>>>>>>>>      >>>>> different environments (e.g. local, custom cluster,
> >>> Vagrant,
> >>>>>>>>>     K8s, Mesos,
> >>>>>>>>>      >>>>> Docker, cloud providers, etc.)
> >>>>>>>>>      >>>>> * Easy to write unit tests for distributed systems
> >>>>>>>>>      >>>>> * Adopted and successfully used by other distributed
> >>>>>>>>> open
> >>>>>>>>>     source project -
> >>>>>>>>>      >>>>> Apache Kafka.
> >>>>>>>>>      >>>>> * Collect results (e.g. logs, console output)
> >>>>>>>>>      >>>>> * Report results (e.g. expected conditions met,
> >>> performance
> >>>>>>>>>     results, etc.)
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> WDYT?
> >>>>>>>>>      >>>>>
> >>>>>>>>>      >>>>> [1] https://github.com/nizhikov/ignite/pull/15
> >>>>>>>>>      >>>>> [2] https://github.com/confluentinc/ducktape
> >>>>>>>>>      >>>>> [3]
> >>>>>> https://ducktape-docs.readthedocs.io/en/latest/run_tests.html
> >>>>>>>>>      >>>>> [4] https://yadi.sk/d/JC8ciJZjrkdndg
> >>>>>>>
> >>>>>>
> >>>>>>
> >>>>>>
> >>>> <2020-07-05--004.tar.gz>
> >>>
> >>>
> >>>
> >>
> >
>