On Wed, Jun 26, 2013 at 11:52 AM, Radu Gheorghe <[email protected]>wrote:
> Thanks for your reply, Rainer! This clears up my options. Now I only have > to decide :) > > I was also thinking about using templates for filenames in imfile. This > requires additional work, though, so it falls in the same category as epoll > support. > > much of this is in the prototype inside the forum. Actually, it went to far in some edges to make it an easy merge. It may pay to do a search. IF you can't find, I can do a search over various sources. And I need to correct myself on a detail: I meant inotify instead of epoll (and I guess you understood me well ;)). Rainer > Best regards, > Radu > > > > 2013/6/26 Rainer Gerhards <[email protected]> > > > On Wed, Jun 26, 2013 at 11:01 AM, Radu Gheorghe <[email protected] > > >wrote: > > > > > Hello, > > > > > > I know the problem of imfile and rotating logs was discussed a million > > > times now, but I still can't find a comprehensive answer. So please > > excuse > > > me for bringing it up again. > > > > > > I understand that previous versions of imfile used to monitor the size > of > > > the log and if that went down, assume it was truncated and start > > monitoring > > > from the beginning. > > > > > > I also understand that this might cause problems if, for example, the > app > > > starts logging like nuts and the size of the log goes up again right > > after > > > truncation. Are there other significant issues that I'm missing? > > > > > > > > I don't remember the exact details, but this was very race. I remember > that > > I spent roughly a month debugging the results, and the only cure out of > > this was disabling this feature, as there was no way to stabilize it > > (something along the lines of read vs. FS metadata vs don't remember ;)). > > > > > > > Assuming the answer is "no", can't the old functionality be offered as > an > > > option (eg: checkFileSize=yes/no)? There can be a disclaimer there and > it > > > could be "no" by default. I think this would be useful not only to me, > > but > > > for other people using Jetty or other apps that don't play nicely with > > > having their logfiles moved for rotating. > > > > > > > > That's probably an option, especially if I put in that turning it on is > > know to cause problems. I still guess that some folks will use this as an > > invitation to file bug reports and especially support customers will > > eventually come up and require assistance with it - something I really > > don't like for a feature I know not to work properly (it's really calling > > for a lot of trouble...). > > > > The real cure, I think, is to rewrite imfile so that it uses epoll, which > > IMHO would remove the core race problem. It's on my list for about 9 > month > > now (3 yrs even as a very low prio item), but I am currently to busy to > > look at it. Would help if it got sponsored. There is even an incomplete > > version of an imfile with epoll support doing by someone else inside the > > forum - if you want to look for that... > > > > Rainer > > > > > Speaking about my situation, here's where I'm at: > > > > > > ########## > > > Context > > > ########## > > > - moving and creating new logfiles makes Jetty stop logging, and > sending > > > HUP to it doesn't do anything. Although this is what imfile seems to > > > prefer, I'd have to restart Jetty, which creates a downtime I can't > > afford > > > - similarly, copying and truncating logfiles confuses rsyslog, do I'd > > have > > > to remove the state file and restart rsyslog. Again, a downtime I can't > > > afford > > > > > > ########## > > > Other possible solutions > > > ########## > > > It seems like there are two solutions I can try. They seem nastier than > > > what I suggest, but might work: > > > - find a logging library for Jetty that plays nicely with the "create" > > > method of rotating. If there is such a library, and I'd have to see > what > > > are the implications for our project > > > - make Jetty log to syslog directly. Again, it's a fuzzy land, because > of > > > the unknown implications, plus I'm not sure how those libraries will > play > > > with large stacktraces (due to size and \n) > > > - use a different rsyslog daemon for monitoring the files, so I can > > afford > > > to restart it > > > > > > Thanks and best regards, > > > Radu > > > _______________________________________________ > > > 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. > > > _______________________________________________ > 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.
