> -----Original Message----- > From: [email protected] [mailto:rsyslog- > [email protected]] On Behalf Of Steffen Sledz > Sent: Tuesday, August 03, 2010 11:06 AM > To: [email protected] > Subject: Re: [rsyslog] [PATCH] break potential infinite loop in > actionDoRetry > > Am 03.08.2010 10:42, schrieb Rainer Gerhards: > > Depends on your environment. If you have a single UDP > > listener, reordering may occur on the wire before you > > even receive it.... > > As you can guess in our context UDP is not involved, just the fifo.
The you have the issue of context switching between producers (assuming there is more than one). > > > Note that when you use a disk assisted queue, the queue > > engine always puts some messages out of order when switching > > to/from Disk mode. Newer builds also consume messages from > > memory and disk concurrently and thus potentially out of > > queue order. > > I think that's what we see. > > So i believe we have to decide between out-of-order messages or loss of > some messages when we do not use a disk assisted queue. :( Or use a disk-only queue. It's performance is obviously very bad, but may be sufficient depending on volume. Rainer _______________________________________________ rsyslog mailing list http://lists.adiscon.net/mailman/listinfo/rsyslog http://www.rsyslog.com
