Igniters,
Currently we have a TeamCity trigger only on Run All aggregating project: https://ci.ignite.apache.org/admin/editTriggers.html?id=buildType:IgniteTests24Java8_RunAll Each change triggers it's own full run. If new changes are added while some builds are queued, these changes are not picked up, and a separate run is scheduled. At this very moment I see over 600 queued builds, all on master branch from some trigger. Previously we had a trigger per-project, which seems to be more efficient. Thoughts? Pavel |
Hi Pavel,
About queue: all these builds were manually triggered by me. It is like experiment under MakeTeamCityGreenAgain. I am trying to accumulate statistics about failing tests in master. Feel free to reoder your build to top. Statistic is necessary to find out how much tests are actually failing. For example in run 1: failed test A in run 2 test A and D, in run 3 B, C As result we actually have A, B, C, D tests failing, but one chain results shows us it is only 1 or 2. I'll share results later. Per project triggers were actually located in template. It was hard to manage. Now Run Basic test chain has VCS trigger (basic&licenses&compute), other builds inlcuded into Run All. Run All does not have VCS triggers and runs every nigth. No one checks results more often then daily. If somethig you may suggest to be included into Run Basic sub-set please share. It should be fast & stable suite without flaky tests because RunBasic now has email notifications. Sincerely, Dmitriy Pavlov пн, 19 февр. 2018 г. в 0:39, Pavel Tupitsyn <[hidden email]>: > Igniters, > > Currently we have a TeamCity trigger only on Run All aggregating project: > > https://ci.ignite.apache.org/admin/editTriggers.html?id=buildType:IgniteTests24Java8_RunAll > > Each change triggers it's own full run. If new changes are added while some > builds are queued, > these changes are not picked up, and a separate run is scheduled. > > At this very moment I see over 600 queued builds, all on master branch from > some trigger. > Previously we had a trigger per-project, which seems to be more efficient. > > Thoughts? > > Pavel > |
In reply to this post by Pavel Tupitsyn
Hi Pavel,
WDYT, 1) is it needed to prepare wiki article describing tests and chains we have, e.g. describe https://ci.ignite.apache.org/viewType.html?buildTypeId=IgniteTests24Java8_RunBasicTests and its difference with RunAll; which chain to use in each case? 2) Which .Net test suite may be applicable to be included into 'run basic tests' (requires <30 min to run, no failed test, no flaky test with fail rate >1%) ? Sincerely, Dmitriy Pavlov пн, 19 февр. 2018 г. в 0:39, Pavel Tupitsyn <[hidden email]>: > Igniters, > > Currently we have a TeamCity trigger only on Run All aggregating project: > > https://ci.ignite.apache.org/admin/editTriggers.html?id=buildType:IgniteTests24Java8_RunAll > > Each change triggers it's own full run. If new changes are added while some > builds are queued, > these changes are not picked up, and a separate run is scheduled. > > At this very moment I see over 600 queued builds, all on master branch from > some trigger. > Previously we had a trigger per-project, which seems to be more efficient. > > Thoughts? > > Pavel > |
Dmitry, thanks for the explanation, this makes sense.
1) Certainly can be helpful. 2) I would include "Platform .NET Core Linux": does not require win agent, runs 15 min, covers most functionality. Pavel On Wed, Feb 21, 2018 at 12:12 AM, Dmitry Pavlov <[hidden email]> wrote: > Hi Pavel, > > WDYT, > 1) is it needed to prepare wiki article describing tests and chains we > have, e.g. describe > https://ci.ignite.apache.org/viewType.html?buildTypeId=IgniteTests24Java8_ > RunBasicTests > and > its difference with RunAll; which chain to use in each case? > > 2) Which .Net test suite may be applicable to be included into 'run basic > tests' (requires <30 min to run, no failed test, no flaky test with fail > rate >1%) ? > > Sincerely, > Dmitriy Pavlov > > пн, 19 февр. 2018 г. в 0:39, Pavel Tupitsyn <[hidden email]>: > > > Igniters, > > > > Currently we have a TeamCity trigger only on Run All aggregating project: > > > > https://ci.ignite.apache.org/admin/editTriggers.html?id= > buildType:IgniteTests24Java8_RunAll > > > > Each change triggers it's own full run. If new changes are added while > some > > builds are queued, > > these changes are not picked up, and a separate run is scheduled. > > > > At this very moment I see over 600 queued builds, all on master branch > from > > some trigger. > > Previously we had a trigger per-project, which seems to be more > efficient. > > > > Thoughts? > > > > Pavel > > > |
Hi Pavel,
Tests are passing now. I've inluded https://ci.ignite.apache.org/viewType.html?buildTypeId=IgniteTests24Java8_IgnitePlatformNetCoreLinux&branch_IgniteTests24Java8=%3Cdefault%3E&tab=buildTypeStatusDiv to Build Basic chain to run suite on per commit basis. Let's see if it tests would be stable Sincerely, Dmitriy Pavlov ср, 21 февр. 2018 г. в 10:26, Pavel Tupitsyn <[hidden email]>: > Dmitry, thanks for the explanation, this makes sense. > > 1) Certainly can be helpful. > 2) I would include "Platform .NET Core Linux": does not require win agent, > runs 15 min, covers most functionality. > > Pavel > > On Wed, Feb 21, 2018 at 12:12 AM, Dmitry Pavlov <[hidden email]> > wrote: > > > Hi Pavel, > > > > WDYT, > > 1) is it needed to prepare wiki article describing tests and chains we > > have, e.g. describe > > > https://ci.ignite.apache.org/viewType.html?buildTypeId=IgniteTests24Java8_ > > RunBasicTests > > and > > its difference with RunAll; which chain to use in each case? > > > > 2) Which .Net test suite may be applicable to be included into 'run basic > > tests' (requires <30 min to run, no failed test, no flaky test with fail > > rate >1%) ? > > > > Sincerely, > > Dmitriy Pavlov > > > > пн, 19 февр. 2018 г. в 0:39, Pavel Tupitsyn <[hidden email]>: > > > > > Igniters, > > > > > > Currently we have a TeamCity trigger only on Run All aggregating > project: > > > > > > https://ci.ignite.apache.org/admin/editTriggers.html?id= > > buildType:IgniteTests24Java8_RunAll > > > > > > Each change triggers it's own full run. If new changes are added while > > some > > > builds are queued, > > > these changes are not picked up, and a separate run is scheduled. > > > > > > At this very moment I see over 600 queued builds, all on master branch > > from > > > some trigger. > > > Previously we had a trigger per-project, which seems to be more > > efficient. > > > > > > Thoughts? > > > > > > Pavel > > > > > > |
In reply to this post by Pavel Tupitsyn
Hi Pavel,
regarding point 1, I've created draft here https://cwiki.apache.org/confluence/display/IGNITE/Continuous+Integration I'll create new topic once it will be ready. SIncerely, Dmitriy Pavlov ср, 21 февр. 2018 г. в 10:26, Pavel Tupitsyn <[hidden email]>: > Dmitry, thanks for the explanation, this makes sense. > > 1) Certainly can be helpful. > 2) I would include "Platform .NET Core Linux": does not require win agent, > runs 15 min, covers most functionality. > > Pavel > > On Wed, Feb 21, 2018 at 12:12 AM, Dmitry Pavlov <[hidden email]> > wrote: > > > Hi Pavel, > > > > WDYT, > > 1) is it needed to prepare wiki article describing tests and chains we > > have, e.g. describe > > > https://ci.ignite.apache.org/viewType.html?buildTypeId=IgniteTests24Java8_ > > RunBasicTests > > and > > its difference with RunAll; which chain to use in each case? > > > > 2) Which .Net test suite may be applicable to be included into 'run basic > > tests' (requires <30 min to run, no failed test, no flaky test with fail > > rate >1%) ? > > > > Sincerely, > > Dmitriy Pavlov > > > > пн, 19 февр. 2018 г. в 0:39, Pavel Tupitsyn <[hidden email]>: > > > > > Igniters, > > > > > > Currently we have a TeamCity trigger only on Run All aggregating > project: > > > > > > https://ci.ignite.apache.org/admin/editTriggers.html?id= > > buildType:IgniteTests24Java8_RunAll > > > > > > Each change triggers it's own full run. If new changes are added while > > some > > > builds are queued, > > > these changes are not picked up, and a separate run is scheduled. > > > > > > At this very moment I see over 600 queued builds, all on master branch > > from > > > some trigger. > > > Previously we had a trigger per-project, which seems to be more > > efficient. > > > > > > Thoughts? > > > > > > Pavel > > > > > > |
Dmitry, awesome start!
Just a note, not every TC project is about JUnit ;) Thanks, Pavel On Fri, Feb 23, 2018 at 12:49 AM, Dmitry Pavlov <[hidden email]> wrote: > Hi Pavel, > > regarding point 1, I've created draft here > https://cwiki.apache.org/confluence/display/IGNITE/Continuous+Integration > > I'll create new topic once it will be ready. > > SIncerely, > Dmitriy Pavlov > > ср, 21 февр. 2018 г. в 10:26, Pavel Tupitsyn <[hidden email]>: > > > Dmitry, thanks for the explanation, this makes sense. > > > > 1) Certainly can be helpful. > > 2) I would include "Platform .NET Core Linux": does not require win > agent, > > runs 15 min, covers most functionality. > > > > Pavel > > > > On Wed, Feb 21, 2018 at 12:12 AM, Dmitry Pavlov <[hidden email]> > > wrote: > > > > > Hi Pavel, > > > > > > WDYT, > > > 1) is it needed to prepare wiki article describing tests and chains we > > > have, e.g. describe > > > > > https://ci.ignite.apache.org/viewType.html?buildTypeId= > IgniteTests24Java8_ > > > RunBasicTests > > > and > > > its difference with RunAll; which chain to use in each case? > > > > > > 2) Which .Net test suite may be applicable to be included into 'run > basic > > > tests' (requires <30 min to run, no failed test, no flaky test with > fail > > > rate >1%) ? > > > > > > Sincerely, > > > Dmitriy Pavlov > > > > > > пн, 19 февр. 2018 г. в 0:39, Pavel Tupitsyn <[hidden email]>: > > > > > > > Igniters, > > > > > > > > Currently we have a TeamCity trigger only on Run All aggregating > > project: > > > > > > > > https://ci.ignite.apache.org/admin/editTriggers.html?id= > > > buildType:IgniteTests24Java8_RunAll > > > > > > > > Each change triggers it's own full run. If new changes are added > while > > > some > > > > builds are queued, > > > > these changes are not picked up, and a separate run is scheduled. > > > > > > > > At this very moment I see over 600 queued builds, all on master > branch > > > from > > > > some trigger. > > > > Previously we had a trigger per-project, which seems to be more > > > efficient. > > > > > > > > Thoughts? > > > > > > > > Pavel > > > > > > > > > > |
Free forum by Nabble | Edit this page |