[
https://issues.apache.org/jira/browse/IMPALA-13669?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17912995#comment-17912995
]
Laszlo Gaal commented on IMPALA-13669:
--------------------------------------
Recent versions of GoogleTest offer two different ways to run death tests: a
fast and a "safe" way. The latter one is suggested for tests that run into
crashes or hangs, at the expense of somewhat slower execution.
See details at
https://github.com/google/googletest/blob/main/docs/advanced.md#death-test-styles.
Changing buffer-pool-test to use the "threadsafe" style of death tests seems to
alleviate the problem, while increasing test duration from ~200 seconds to ~
250 s (on Ubuntu 20.04), which seems acceptable.
> buffer-pool-test hangs on Rocky 9
> ---------------------------------
>
> Key: IMPALA-13669
> URL: https://issues.apache.org/jira/browse/IMPALA-13669
> Project: IMPALA
> Issue Type: Bug
> Components: Backend
> Affects Versions: Impala 4.5.0
> Reporter: Laszlo Gaal
> Assignee: Laszlo Gaal
> Priority: Critical
>
> Recent test runs on Rocky Linux 9.2 often resulted in an hang in
> {{buffer-pool-test}} during BE tests. The hangs were observed only on Rocky
> 9, and they were seen on Intel and ARM CPUs both.
> When the hang occurs, it is only resolved by the test run's internal watchdog
> timing out at 20 hours, killing the build.
> Example runs:
> * https://jenkins.impala.io/job/rocky-9.2-from-scratch-ARM/4/ (ARM)
> * https://jenkins.impala.io/job/rocky-9.2-from-scratch/9/ (Intel)
> Multiple occurrences were observed in private environments as well.
> Marking as P2 (critical), as it doesn't block precommit runs, but makes it
> impossible to make progress with Rocky 9 / RHEL 9 support.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
