[DISCUSSION] Further development ignitevisorcmd

[Nikolai Kulagin]

Hello, Igniters!

I would like to raise a question on the work of the ignitevisorcmd.
Currently, part of the exceptions in the ignitevisorcmd is not thrown and
displayed on the console as WARNING. Example:
VisorOpenCommand#open(String).

The result of this is:
1. Complicated perception of an exception by the user. It seems to me that
if a fat client throws an exception, the ignitevisorcmd should show this as
an exception (or ERROR).
2. Most tests in the ignitevisorcmd do not have a final verification of the
conditions. For example, the VisorKillCommandSpec test is successful,
although the ignitevisorcmd cannot even connect to the cluster. The test
falls only in case of an exception, but some of the exceptions are not
thrown and are displayed as WARNING. Paradox.

And it seems to me that there are two ways to solve the problem:
1. We throw exceptions. In this case, some of the tests begin to work;
   or
2. We continue to display exceptions like WARNING (or ERROR). Then we need
to do the following:
- Correct the tests so that they check the fulfillment of the conditions;
- It might be worth displaying exceptions as an ERROR.

A good example is IGNITE-12757. All tests pass successfully, although the
ignitevisorcmd on default configs does not even connect to the cluster.

WDYT?

[1] https://issues.apache.org/jira/browse/IGNITE-12757

[dmagda]

Hi Nikolay,

I think that your proposal is reasonable. Agree, that all exceptions need
to be signaled properly - either with an ERROR message or by failing an
operation initiated by Visor CMD (you can return some error code and the
tool needs to process it properly). It's up to you and other Visor
maintainers to decide what's the best way to handle such exceptions.

-
Denis


On Tue, Mar 17, 2020 at 12:13 AM Nikolai Kulagin <[hidden email]>
wrote:

> Hello, Igniters!
>
> I would like to raise a question on the work of the ignitevisorcmd.
> Currently, part of the exceptions in the ignitevisorcmd is not thrown and
> displayed on the console as WARNING. Example:
> VisorOpenCommand#open(String).
>
> The result of this is:
> 1. Complicated perception of an exception by the user. It seems to me that
> if a fat client throws an exception, the ignitevisorcmd should show this as
> an exception (or ERROR).
> 2. Most tests in the ignitevisorcmd do not have a final verification of the
> conditions. For example, the VisorKillCommandSpec test is successful,
> although the ignitevisorcmd cannot even connect to the cluster. The test
> falls only in case of an exception, but some of the exceptions are not
> thrown and are displayed as WARNING. Paradox.
>
> And it seems to me that there are two ways to solve the problem:
> 1. We throw exceptions. In this case, some of the tests begin to work;
>    or
> 2. We continue to display exceptions like WARNING (or ERROR). Then we need
> to do the following:
> - Correct the tests so that they check the fulfillment of the conditions;
> - It might be worth displaying exceptions as an ERROR.
>
> A good example is IGNITE-12757. All tests pass successfully, although the
> ignitevisorcmd on default configs does not even connect to the cluster.
>
> WDYT?
>
> [1] https://issues.apache.org/jira/browse/IGNITE-12757
>

[Ilya Kasnacheev]

Hello!

I recommend filing issues and trying to tackle them yourself, since you
seem to be a rare active user of ignitevisorcmd.

Then again, maybe we should move more functionality to control.sh.

Regards,
--
Ilya Kasnacheev


вт, 17 мар. 2020 г. в 10:13, Nikolai Kulagin <[hidden email]>:

> Hello, Igniters!
>
> I would like to raise a question on the work of the ignitevisorcmd.
> Currently, part of the exceptions in the ignitevisorcmd is not thrown and
> displayed on the console as WARNING. Example:
> VisorOpenCommand#open(String).
>
> The result of this is:
> 1. Complicated perception of an exception by the user. It seems to me that
> if a fat client throws an exception, the ignitevisorcmd should show this as
> an exception (or ERROR).
> 2. Most tests in the ignitevisorcmd do not have a final verification of the
> conditions. For example, the VisorKillCommandSpec test is successful,
> although the ignitevisorcmd cannot even connect to the cluster. The test
> falls only in case of an exception, but some of the exceptions are not
> thrown and are displayed as WARNING. Paradox.
>
> And it seems to me that there are two ways to solve the problem:
> 1. We throw exceptions. In this case, some of the tests begin to work;
>    or
> 2. We continue to display exceptions like WARNING (or ERROR). Then we need
> to do the following:
> - Correct the tests so that they check the fulfillment of the conditions;
> - It might be worth displaying exceptions as an ERROR.
>
> A good example is IGNITE-12757. All tests pass successfully, although the
> ignitevisorcmd on default configs does not even connect to the cluster.
>
> WDYT?
>
> [1] https://issues.apache.org/jira/browse/IGNITE-12757
>