28 May
2008
28 May
'08
1:23 a.m.
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/