Loggregator not getting everything crashlogs used to get

36 views
Skip to first unread message

Mike Youngstrom

unread,
Jul 1, 2014, 3:58:12 PM7/1/14
to vcap...@cloudfoundry.org
I've had some trouble lately debugging application crashes using loggregator.  I realized that crashlogs would have given me more data to help with my diagnosis.

I'm running 173.

Test Case:

Push an application with a command that you know will fail.

gcf push test-crash -c "bad command"

After it fails use the old ruby cli or hit the crashlogs api directly.  Look at stderr.log.  I see:

cf crashlogs test-crash

stderr.log:
"/bin/bash: line 26: bad: command not found"

If you look in loggregator you don't see this error anywhere.

Thoughts?    Perhaps loggregator isn't attaching to an apps stderr fast enough?  Is this a known issue?

Mike

Alex Jackson

unread,
Jul 2, 2014, 9:24:33 AM7/2/14
to vcap...@cloudfoundry.org
Hi Mike,
    Yes there can be a slight delay before loggregator attaches to the applications stdout / stderr.    I'm not sure that this is causing the issue you seeing.   The error is occurring during application staging correct?   The DEA might not be sending all the output it could about application staging to loggregator.   I can look over the code and see if that's the case.    

I do know that in diego, loggregator is more of a first class component.   The design with diego and loggregator should remove the "race conditions" around application staging / startup and loggregator data capture.
       - Alex.


--
You received this message because you are subscribed to the Google Groups "Cloud Foundry Developers" group.
To view this discussion on the web visit https://groups.google.com/a/cloudfoundry.org/d/msgid/vcap-dev/CAEoPEDqWSn0fGtHN6yq5X3Hth%3Dxat-C2OeGqWO0-et8HqjLimw%40mail.gmail.com.

To unsubscribe from this group and stop receiving emails from it, send an email to vcap-dev+u...@cloudfoundry.org.

Mike Youngstrom

unread,
Jul 2, 2014, 11:51:53 AM7/2/14
to vcap...@cloudfoundry.org
Sorry I wasn't clear.  My example skipped staging.  So it was an application startup error not getting logged in loggregator.

It sounds like this is a known issue though.  So, I'll try it again with Diego.

Thanks,
Mike


Reply all
Reply to author
Forward
0 new messages