On Fri, Jul 12, 2013 at 12:01 PM, Jonny Törnbom <[email protected]>wrote:
> On Fri, Jul 12, 2013 at 11:37:32AM +0200, Rainer Gerhards wrote: > > On Fri, Jul 12, 2013 at 11:44 AM, Jonny Törnbom <[email protected] > >wrote: > > > > > On Fri, Jul 12, 2013 at 11:11:04AM +0200, Rainer Gerhards wrote: > > > > On Fri, Jul 12, 2013 at 11:03 AM, Jonny Törnbom < > [email protected] > > > >wrote: > > > > > > I agree that it is really a problem outside of rsyslog, even if it > > > affects rsyslog (or any other logging solution for that matter). We are > > > bugging the "other mailing list" about the issue and will hopefully > come > > > up with a solution. > > > > > > Now I know atleast that there isn't an easy workaround from rsyslog > > > side, but I will definitely let you know if what we come up with can > > > be of help to other rsyslog+systemd users. > > > > > > > > Well, technically the problem is that the issue already has happened > before > > rsyslog is even started. So whatever we would try to do (e.g. increasing > > buffers) doesn't help, as it is too late at that point in time. So I > don't > > see any solution except that the journal gets fixed. If I would see one, > > I'd definitely try to implement it, but again: whatever we do -- it's > > simply too late, damage already occured... > > > > Rainer > > The problem occurs before any logging daemon starts, it is not the fault > of rsyslog or any other logging daemon, indeed. What I meant as > workaround from an rsyslog perspective would perhaps be something like > starting rsyslog really really early in a systemd environment. > > OK... This is really nothing that the rsyslog project wants to mangle with. We think it would be a bad idea if we specify startup order. We don't know the specifics of all distros (and frankly do not want to). So IMHO it is the distro maintainer's part to decide when is the right time to start rsyslog (depends on what becomes writable when, which services are used by default [databases etc] and so on). The confusing part may be that our git carries some [contributed, non-project] systemd control files. They are meant as a convenience to whoever may do the integration. It's nothing that we actively maintain (for the reason given above). > With solutions I meant solutions on the "other side", systemd fixes that > affects any syslog daemon starting up. If such a "generic" solution > comes up, I'll notify you about it since it could help other > syslog users (in this case rsyslog users) running systemd. :-) > > yeah, I didn't make myself clear enough, sorry for that. I understood you correctly, just wanted to explain that I really don't see an option to change things from the rsyslog side. If I were a journald developer, I would check if at least I could increase the socket buffers size. The proper solution of course is to buffer things until the socket consumer is up and running. Hope that clarifies, Rainer > Regards, > Jonny > _______________________________________________ > 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. > _______________________________________________ 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.
