Squirrel wrote: > Hi guys, > > Some days ago may server got some high CPU load and was unavailable to any > web-request. When I checked what causes it I found out it was Tomcat5.5. So > I did run some benchmark tests with ab (Apache Benchmark) and found out that > I can repeatedly crash Tomcat5.5 with some 'high' concurrent values (ab -n > 400 -c 100 http://domain.com/xwiki/bin/view/Main/). > > In some posts it was hinted that not Tomcat itself crashs, but the webapp > itself and the developers may be able to track what causes the loop. > > Are you able to reproduce that? Is it really a XWiki issue or is it indeed a > Java/Tomcat issue? Can I help in someway? > > Cheers >
It is not an XWiki issue, but there are some issues in the used libraries. For example, Apache FOP enters in a neverending loop when exporting the XWiki.XWikiSyntax page (this is fixed in fop-0.95, which will be used in XWiki 1.5), and OSCache has some deadlocks under high load (this is why we're trying to switch to other cache providers, like memcached or JBoss, which will also be part of 1.5). I've been using XWiki on 3 medium-load servers (one of them has a traffic ranking of about 50.000 according to alexa.com), and so far I didn't encounter any problems. The server may be slow, but it all depends on the settings, like the size of the cache, the number of allowed database connections, proper skin files caching, caching the rendering result for documents that take a lot of time (using $context.setCacheDuration(<number of seconds>)), tweaking queries, and so on. -- Sergiu Dumitriu http://purl.org/net/sergiu/ _______________________________________________ devs mailing list [email protected] http://lists.xwiki.org/mailman/listinfo/devs
