lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Dawid Weiss (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (LUCENE-4510) when a test's heart beats it should also throw up (dump stack of all threads)
Date Sat, 27 Oct 2012 07:41:13 GMT

    [ https://issues.apache.org/jira/browse/LUCENE-4510?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13485376#comment-13485376
] 

Dawid Weiss commented on LUCENE-4510:
-------------------------------------

I vaguely remember people on the mailing list expressed concerns even about the heartbeat
messages so throwing stack traces at everyone would be probably a bit too much. I see a few
options.

1) Stack traces are dumped at test/suite timeout, so you could simply override the default
and lower this timeout locally:

For tests only:
-Dtests.timeout=5000!

For entire suites:
-Dtests.timeoutSuite=10000!

Note the '!' at the end which means the property should take precedence over annotations.
The scale is millis.

2) The above workaround has an obvious problem that the stack trace is not dumped repeatedly
which can be extremely useful in telling  which parts of the code a thread was locked in.
In fact, randomized runner once HAD a snippet of code which was doing it automatically --
on timeout it would probe stack traces of all threads a few times at random intervals and
then it would print the 'diverging' paths for each thread (along with the constant part).
I removed it because it wasn't used by anybody :)

So... I could add yet another option that would dump stack traces on heartbeats but it would
be turned off by default which I think is sensible. This option could also take a number which
would mean roughly: "start dumping stack traces after the N-th heartbeat".

Let me know if (2) sounds good to you.

                
> when a test's heart beats it should also throw up (dump stack of all threads)
> -----------------------------------------------------------------------------
>
>                 Key: LUCENE-4510
>                 URL: https://issues.apache.org/jira/browse/LUCENE-4510
>             Project: Lucene - Core
>          Issue Type: Bug
>            Reporter: Michael McCandless
>
> We've had numerous cases where tests were hung but the "operator" of that particular
Jenkins instance struggles to properly get a stack dump for all threads and eg accidentally
kills the process instead (rather awful that the same powerful tool "kill" can be used to
get stack traces and to destroy the process...).
> Is there some way the test infra could do this for us, eg when it prints the HEARTBEAT
message?

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org


Mime
View raw message