queue-persist.sh test sometimes fail

[Whissi]

The queue-persist.sh test randomly fail with

FAIL: queue-persist.sh
======================

[queue-persist.sh]:
testing memory queue persisting to disk, mode LinkedList
rsyslogd started with pid  105078
injecting 5000 messages
imdiag[13500]: 5000 messages injected
There must exist some files now:
total 2172
-rw------- 1 portage portage 1048581 Feb  8 16:35 mainq.00000001
-rw------- 1 portage portage 1048581 Feb  8 16:35 mainq.00000002
-rw------- 1 portage portage  111718 Feb  8 16:35 mainq.00000003
-rw------- 1 portage portage     485 Feb  8 16:35 mainq.qi
rsyslogd started with pid  105099
imdiag[13500]: mainqueue empty
read value 15, but expected value 14
sequence error detected
FAIL queue-persist.sh (exit status: 1)

Discussion on mailing list: http://thread.gmane.org/gmane.comp.sysutils.rsyslog/16401

[alorbach]

FYI: This seems to be a bug in queue processing on shutdown introduced in V8, V7 stable doesn't have this problem.

[rgerhards]

Not sure on the "small" label. May be a complex problem during queue shutdown. I have now tracked this issue on several testbench platforms. It may simply be related to testbench timing, but it may also point into a weired condition that can also fail in practice. We should investigate this further under this assumption.

I will also try to disable this test for now, so that we can see if it is the remaining one that causes instability of testbench results.

[rgerhards]

relates to: #604

[rgerhards]

I think I finally found a grip on what is going on here. The good thing is that I am now able to more or less consistenly reproduce within a couple of tries. The bad news is that this is not related to testbench timing, but to queue shutdown, as Andre suggested. So it seems to showcase a real problem, which might cause the loss of exactly one message during queue shutdown when special timing happens. I am further investigating but thought I share the update.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.