> Exactly. My main thought error, as I currently see it, is that while it is > great to > delay inputs when they submit to main queues, it is counter-productive to try > the > same with action queues. After all, the messages are already accepted and we > would actually delay the main q worker (this is exactly what happens). I guess > this is at the root of the problems we currently see. But I need to do some > more > analysis...
Yes, that was exactly the problem. A patch is available here: http://git.adiscon.com/?p=rsyslog.git;a=commitdiff;h=fd46d10a4f49da35758ab9201f6f863d69d5d4f5 It works for me, at least during quick testing. Would be great if you could also try it out. Note, however, that it does not (and cannot) solve the issue that a filled-up action message queue will cause delays on the main queue. You can reproduce this problem if you remove the disk queue and let your script run. There is no way to prevent this without discarding messages in the action queue. However, I need to check if the action queue is configurable enough so that one can configure immediate discarding on queue full state. I think that's currently not possible. 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
