On Thu, Nov 7, 2013 at 10:01 AM, Rainer Gerhards
<[email protected]>wrote:

> On Wed, Nov 6, 2013 at 9:33 PM, Pavel Levshin <[email protected]>wrote:
>
>> wtpAdviseMaxWorkers() is called from several places, but most of them are
>> related to DA shutdown. One place where it is used for all workers is
>> qqueueAdviseMaxWorkers(). And there, it is called separately for regular
>> and DA pool. It should be OK, but I feel there is a bug, which may, under
>> rare circumstances, lead to stall with my patch. It is probably a bug even
>> without the patch.
>>
>
> I've also worked a little bit more on this  and you are right, there is
> something fishy. I got some (of many) testbench runs which did not work
> correctly on queue-related things. It looks like this is an older
> regression. I'll dig deeper in that and will also look closely at what you
> say/propose below. It may take a short while as what I see does not
> consistently appear. I am not sure that what you describe below actually
> *is* the issue -- it looks more like some misadressing somewhere to me (I
> dont say I disagree to your analysis, just haven't checked it in depth yet
> as I was wroking on the testbench unti now).
>
> I will keep you posted.
>

Ok, the fishyness I saw is rooted in --enable-rtinst, where a race occurs
at rsyslog termination. I haven't used it for a while, and it's even
questionable if it still is useful (it's rooted in far less capable
debugging tools years ago...). As it is to be enabled only for testing, I
ignore the problem for a moment and just keep it disabled. But I have made
sure it actually is the root cause of what I have seen.

I'll now check back on the threading patch.

Rainer
_______________________________________________
rsyslog mailing list
http://lists.adiscon.net/mailman/listinfo/rsyslog
http://www.rsyslog.com/professional-services/
What's up with rsyslog? Follow https://twitter.com/rgerhards
NOTE WELL: This is a PUBLIC mailing list, posts are ARCHIVED by a myriad of 
sites beyond our control. PLEASE UNSUBSCRIBE and DO NOT POST if you DON'T LIKE 
THAT.

Reply via email to