[
https://issues.apache.org/jira/browse/CASSANDRA-18665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17773238#comment-17773238
]
Maxim Muzafarov edited comment on CASSANDRA-18665 at 10/9/23 10:51 AM:
-----------------------------------------------------------------------
{quote}
Isn't bound to one directory, the call to generatetokens returning an empty
list is usually the first sign. After that there is various errors like "Could
not determine current working directory", "Cannot find 'logs' directory", and
cannot invoke 'java -version'.
{quote}
I think I've managed to fix the problem here (note, that this is not the
minimum required solution and is not optimal, from my point of view):
https://ci-cassandra.apache.org/job/Cassandra-intree-5.0-dtest-upgrade/50/
I've found that each matrix job checkouts the cassandra project dir to the
{split} directory, but the cassandra-dtest is shared among the running jobs.
So, when a matrix job finishes, it deletes the shared cassandra-dtest
directory, which, in turn, affects the other running jobs on the same agent.
So, here's how I fixed it (a number is the number of split):
{code}
drwxr-xr-x 19 jenkins jenkins 4096 Oct 8 12:49 3
drwxr-xr-x 19 jenkins jenkins 4096 Oct 8 13:42 30
drwxr-xr-x 19 jenkins jenkins 4096 Oct 8 12:46 40
drwxr-xr-x 19 jenkins jenkins 8192 Oct 8 12:49 cassandra-dtest3
drwxr-xr-x 19 jenkins jenkins 8192 Oct 8 13:41 cassandra-dtest30
drwxr-xr-x 19 jenkins jenkins 8192 Oct 8 12:45 cassandra-dtest40
{code}
This seems to me a sub-optimal solution. From my point of view, I think it is
better to have cassandra and cassandra-dtest for each split independently to
avoid such problems in the future, even if we need to spend some extra disk
space on it. For example, we can put Cassandra sources under
{{{split}/cassandra}} (they are now in the {split} directory) and the Cassandra
dtest sources in {{{split}/cassandra-dtest}}, so that we will not have the same
problems in the future and reach a good level of isolation.
was (Author: mmuzaf):
{quote}
Isn't bound to one directory, the call to generatetokens returning an empty
list is usually the first sign. After that there is various errors like "Could
not determine current working directory", "Cannot find 'logs' directory", and
cannot invoke 'java -version'.
{quote}
I think I've managed to fix the problem here (note, that this is not the
minimum required solution and is not optimal, from my point of view):
https://ci-cassandra.apache.org/job/Cassandra-intree-5.0-dtest-upgrade/50/
I've found that each matrix job checkouts the cassandra project dir to the
{split} directory, but the cassandra-dtest is shared among the running jobs.
So, when a matrix job finishes, it deletes the shared cassandra-dtest
directory, which, in turn, affects the other running jobs on the same agent.
So, here's how I fixed it (a number is the number of split):
{code}
drwxr-xr-x 19 jenkins jenkins 4096 Oct 8 12:49 3
drwxr-xr-x 19 jenkins jenkins 4096 Oct 8 13:42 30
drwxr-xr-x 19 jenkins jenkins 4096 Oct 8 12:46 40
drwxr-xr-x 19 jenkins jenkins 8192 Oct 8 12:49 cassandra-dtest3
drwxr-xr-x 19 jenkins jenkins 8192 Oct 8 13:41 cassandra-dtest30
drwxr-xr-x 19 jenkins jenkins 8192 Oct 8 12:45 cassandra-dtest40
{code}
This seems to me a sub-optimal solution. From my point of view, I think it is
better to have `cassandra` and `cassandra-dtest` for each split independently
to avoid such problems in the future, even if we need to spend some extra disk
space on it. For example, we can put Cassandra sources under {split}/cassandra
(they are now in the {split} directory) and the Cassandra dtest sources in
{split}/cassandra-dtest, so that we will not have the same problems in the
future and reach a good level of isolation.
> Update jenkins groovy dsl to use in-tree scripts
> ------------------------------------------------
>
> Key: CASSANDRA-18665
> URL: https://issues.apache.org/jira/browse/CASSANDRA-18665
> Project: Cassandra
> Issue Type: Task
> Components: CI
> Reporter: Michael Semb Wever
> Assignee: Michael Semb Wever
> Priority: Normal
> Fix For: 5.0.x, 5.x
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
