Spent only a few minutes looking at this, so don't have an in-depth analysis, but here is what I observed. It seems most accept/close calls are done from thread 40210 and 40211. The pattern is, it seems to have a 138 socket buffer, it fills it up accepting connections in anger (the only thing it does after accepting connections is setting of flags (read_write + no_block)). It accumulates connections like this until it has 138, then goes about reading-from and closing a few. Now one of the problems could be, it doesn't get to close all of them. For instance, for thread 40210, it started at FD 11 the first time, ended at 145, then after the epoll returning EAGAIN, it closed a few and then went about accepting new ones (while working on reading/closing existing ones along the way). It again started at 11 (but some sockets weren't read/closed) and it finished at 162(afair). So it basically has some 20 odd sockets still opened, lying around. But then this count never comes down, next time it ends at 164 (or 166, maybe). The end-socket count never reached below 162 after that.
Now, if eventually the socket is closed, it would have no problem, but if it isn't, its eventually is going to lead upto 1024 open FDs (default open fd limit in most Linux environments). The accept calls are all for FD 4 (which is bound to port 13514). The threads are forked from 40206 (rsyslogd) process. On Fri, Nov 14, 2014 at 8:31 PM, Thomas D. <[email protected]> wrote: > Hi, > > is https://www.dropbox.com/s/dapbxs4xtqlwmu1/rsyslog1550.log.gz?dl=1 > better? > > Test log from this run: https://bpaste.net/show/7f403855e381 > > > One thing I noticed: Every time I run with valgrind or strace, it is not > failing with > > > libgcc_s.so.1 must be installed for pthread_cancel to work > > At least I don't see this error message somewhere. > > From the test log it is failing with > > > sequence error detected > > > -Thomas > _______________________________________________ > 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. > -- Regards, Janmejay http://codehunk.wordpress.com _______________________________________________ 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.
