lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From François Schiettecatte <fschietteca...@gmail.com>
Subject Re: Garbage collection issue and RELOADing cores
Date Tue, 01 Jul 2014 12:36:17 GMT
Hi

Just following up on my previous post about a memory leak when RELOADing cores, I narrowed
it down to the SuggestComponent, specifically '<searchComponent name="suggest" class="solr.SuggestComponent">...</searchComponent>'
in solrconfig.xml. Comment that out and the leak goes away.

The leak occurs in 4.7, 4.8 and 4.9. It occurs when a core is RELOADed, but not if it is UNLOADed
and then LOADed. It occurs whether G1, CMS or ParallelGC is used for garbage collection.

I used JDK 1.7.0_60 and Tomcat 7.0.54 for the underlying layers.

Not sure where to take it from here?

Cheers

François


On Jun 16, 2014, at 4:50 PM, François Schiettecatte <fschiettecatte@gmail.com> wrote:

> Hi
> 
> I am running into an interesting garbage collection issue and am looking for suggestions/thoughts.

> 
> Because some word lists such as synonyms, plurals, protected words need to be updated
on a regular basis I have to RELOAD a number of cores in order to 'pick up' the new lists.

> 
> What I have found is that I get a memory leak when I do a RELOAD rather than an UNLOAD/CREATE
with core admin. This is most pronounced with the G1 GC and much less so with the CMS GC.
The former will cause the VM to run out of memory after 5/6 RELOADs, while the latter does
so after 30/35 RELOADs. We are not talking about large indices here, the files footprint totals
470MB.
> 
> I am using SOLR 4.8.1, Tomcat 7.0.53, jdk1.7.0_60, on Fedora Core 20. I am not using
any fancy GC parameters, I cut everything back to basics, just:
> 
> 	-Xmx1G -XX:+UseConcMarkSweepGC -XX:+UseParNewGC
> 
> and 
> 
> 	-Xmx1G -XX:+UseG1GC
> 
> I was curious if anyone else had run into this issue and managed to fix it?
> 
> Thanks
> 
> François
> 
> 
> 


Mime
View raw message