[opencms-dev] Tomcat 5.0.16 usage

[Mariano Barcia]

Matt,

The problems were strictly related to the indexation process with
Lucene. The JVM would not stop until it reached 1GB RAM memory usage,
and the CPU would go nuts when the usage had reached the JVM max heap
limit. 

Actually, I must say the problem was not about performance except for
the special case mentioned above for the CPU.

The good news is that the customer performed stress tests on the system,
with great success even with about 2GB of contents loaded, so don't
worry about the module, it works fine ;-) great work! So far, with more
than 4000 html documents uploaded and indexed, the search engine works
great, simply unbeatable.

We are using v1.4 of the opencms-lucene module, with some modifications
Ernesto did, but now I believe they are integrated into the last v1.5
release, we could use v1.5 right?

Thanks,
--mariano

PS: We could issue a .class file to patch Opencms 5 for Xerces v2.5,
would you host it on http://opencmslucene.sourceforge.net ?


-----Mensaje original-----
De: opencms-dev-admin at opencms.org [mailto:opencms-dev-admin at opencms.org]
En nombre de M Butcher
Enviado el: Viernes, 06 de Febrero de 2004 08:15 p.m.
Para: opencms-dev at opencms.org
Asunto: Re: [opencms-dev] Tomcat 5.0.16 usage

Mariano

Mariano Barcia wrote:
> Reed, Matt, all,
> 
> I'd like to share a experience we had recently with a opencms website
> using Lucene as the search engine.

Were the problems related directly to Lucene? (Could you tell?) The 
Lucene module does nor really use Xerces much, relying on the OpenCms 
API for as much as possible. However, I suppose a high volume of search 
requests could cause some slowing if the registry.xml file was being 
repeatedly parsed.

If the poor performance is directly related to the Lucene module, we'll 
see if we can do some performance tuning or add caching to speed things 
up. BTW -- which version of the Lucene module are you running?

> We discovered that Xerces v1.4.4 has serious performance and memory
> usage problems with Opencms-Lucene, which were solved when replaced by
> v2.5. Xerces v1.4.4 is the library bundled with OpenCMS v5.0.x, but so
> far the only non-2.5-compliant method we've seen is the one mentioned
> here by Reed (getOriginalEncoding). We changed that method to comply
> with v2.5 (easy, just change "getEncoding" for "getXmlEncoding") and
> everything has been working from then.
> 
> We have not had time yet to check on the CVS HEAD version of Opencms,
> maybe Opencms 5.3.1 has Xerces v2.x now ?

You should send them a patch against 5.0.1! Some of us still build out 
of the CVS release, and I suppose there is also the possibility of a 
5.0.2 release. If that's the only thing that is stopping newer versions 
of Xerces from working, then I'd rather see a fix very soon. I'm getting

tired of seeing the same Xerces posting repeatedly appear on the list.
;-)

Thanks,

Matt
_______________________________________________
This mail is send to you from the opencms-dev mailing list
To change your list options, or to unsubscribe from the list, please
visit
http://mail.opencms.org/mailman/listinfo/opencms-dev