On 7 August 2013 21:14, Raphael Bircher <r.birc...@gmx.ch> wrote: > Am 07.08.13 21:04, schrieb Rory O'Farrell: > > On Wed, 7 Aug 2013 20:35:30 +0200 >> janI <j...@apache.org> wrote: >> >> On 7 August 2013 17:28, Ricardo Berlasso <rgb.m...@gmail.com> wrote: >>> >>> 2013/8/7 FR web forum <ooofo...@free.fr> >>>> >>>> I pull up this issue. >>>>> The error occurs again since two weeks. >>>>> >>>>> >>>>> Today, forums alternate between error 1040 and being slower than a >>>> snail. >>>> It is not possible to access them right now. >>>> >>>> I think you have some fast snails in your area :-) >>> >>> At least the vm itself seems relative ok now. >>> >>> However, I can see that there has been mysql problems earlier, seems to >>> be >>> the old problem, that httpd tries to grap more connections than >>> available. >>> >>> This time I could also see that the mysql buffer spaces had been pretty >>> full, without knowing the forum software, I would say that some of the >>> tables should be converted to innodb. We did that on mwiki, and that >>> change >>> alone allowed mysql to handle a lot more parallel select statements >>> (60-70% >>> of the total). >>> >>> In general the configs are ok for normal daily traffic but cannot handle >>> peak situations. >>> >>> Rgds >>> jan I. >>> >>> >>> Regards >>>> Ricardo >>>> >>>> >>>> >>>> For interested look at: >>>>>> INFRA-6097 which request mysql optimization and >>>>>> INFRA-6098 which request httpd/ats/php optimization >>>>>> They are assigned to me, but when I do it, I need somebody to help >>>>>> test...any volunteers ? >>>>>> >>>>> Just before one outage of the forum today there were 150 users online, >> which is a long way short of the maximum logged of 294. An interesting >> thing I have noted is that it is often slow, leading to an error 10040 at >> about 0930 UTC+1; this happens quite regularly. >> >> Maybe this is the backup. Bugzilla has the same behavure at the same > time. Or ASF Server simply don't like morning ;-) >
good tip to the forum adminidtrator, backup script start at 6:46 UTC, with a sqldump, then generating/encrypting backup files and finally rsync to the backup server. This costs some cpu cycles and takes memory. This can cause a simualted peak situation, when some users (30-40) at the same time are quite active then the vm will have a problem. rgds jan I. > > Greetings Raphael > > > ------------------------------**------------------------------**--------- > To unsubscribe, e-mail: > dev-unsubscribe@openoffice.**apache.org<dev-unsubscr...@openoffice.apache.org> > For additional commands, e-mail: dev-h...@openoffice.apache.org > >
