Hi, We have some segmentation faults reported for children but these don't bring the server down or cause any problems at all and they do not relate to the problem described earlier. The Processors that get stuck do not segfault when they die they simply get killed by ExpireTimeout.
We run CentOS 5.5 on a recent Linux 64bit kernel. Thanks. On 4 Apr 2011, at 08:17, Michal Kumžák wrote: > Hello > > Do you have a problem with segmentation fault on your server? Wath OS is on > your server? > Thanx > > Michal Kumzak > > Dne 4.4.2011 2:17, Andrew napsal(a): >> Hi, >> >> We are using the latest v16 peruser testing patch. We have noticed that >> IdleTimeout is not getting applied to Processors. >> >> To see this restart peruser and then visit a site you host. We see on the >> server-status page that peruser spawns 2 processes: one is called Processor >> and the other Worker. The Worker process gets killed after IdleTimeout but >> the Processor process sits in a "Ready" state until ExpireTimeout is >> reached. This stuck Processor bizarrely reports itself as both active *and* >> idle on the server-status page for the whole time it's sitting there after >> having served the initial (and only) request for the domain! >> >> As I understand it ExpireTimeout is a safety rule and very specifically >> IdleTimeout should be applying to all Processors that sit idle. We have a >> high ExpireTimeout set (for good reasons as we don't want to interrupt >> people downloading large files) but a short IdleTimeout as we want to >> conserve memory on our 3000+ VirtualHost setup and as it stands having these >> idle Processor processes hanging about until ExpireTimeout is reached is >> causing us problems. >> >> Here is our peruser configuration: >> >> ServerLimit 5000 >> MaxClients 1000 >> MinSpareProcessors 0 >> MaxSpareProcessors 5 >> MinProcessors 0 >> MaxProcessors 50 >> MaxRequestsPerChild 1000 >> ExpireTimeout 600 >> IdleTimeout 30 >> MultiplexerIdleTimeout 120 >> ProcessorWaitTimeout 5 10 >> MinMultiplexers 5 >> MaxMultiplexers 250 >> >> Our<Processor> directives simply contain User, Group and Chroot >> configurations. >> >> Any advice appreciated! Perhaps this is a bug or have we missed a >> configuration directive such as IdleProcessorTimeout? >> >> Thanks. >> _______________________________________________ >> Peruser mailing list >> [email protected] >> http://www.telana.com/mailman/listinfo/peruser > _______________________________________________ > Peruser mailing list > [email protected] > http://www.telana.com/mailman/listinfo/peruser _______________________________________________ Peruser mailing list [email protected] http://www.telana.com/mailman/listinfo/peruser
