lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Shawn Heisey <s...@elyograg.org>
Subject Re: "Classic" 4.2 master-slave replication not completing
Date Wed, 03 Jul 2013 16:28:21 GMT
On 7/3/2013 9:29 AM, Neal Ensor wrote:
> Posted the solr config up as http://apaste.info/4eKC (hope that works).
>   Note that this is largely a hold-over from upgrades of previous solr
> versions, there may be lots of cruft left over.  If it's advisable to do
> so, I would certainly be open to starting from scratch with a 4.3+ example
> configuration.
>
> I will take up the logging portions you listed there, to see where I might
> have missed a library or configuration point, thanks.

You have a LOT of auto warming in your cache configuration.  Unless your 
filter queries are really super simple, a more realistic autowarmCount 
for the filterCache is low double digits or even single digit.  If they 
are super simple, a number like 64 might be OK.  For the 
queryResultCache, I don't think I would go above a few hundred at the 
absolute most.  You probably don't need overall cache sizes that large, 
but that shouldn't cause any problems, unless it is pushing the limits 
of your java heap.

The high numbers that you have mean that every commit with openSearcher 
set to true (the default) will result in Solr running 4608 queries 
against the new index before it can go online.  If both master and slave 
servers have similar config, it will happen on both.  As you might 
imagine, this many queries can take a very long time, which can lead to 
a whole cascade of other problems, and I wouldn't be surprised to learn 
that replication is affected.

Try reducing the autowarmCount values drastically and see what happens.

Thanks,
Shawn


Mime
View raw message