Thanks for the response. I double-checked that we don't have the core open
multiple times. The complete index size is about 200M (around 1,060,000
documents). During the indexing process, 26 files were created. Core admin
interface indicated that no query or process were running after roughly 5 hours
but the "Time Elapsed" was still going. We have the indexDefults setting as
followed:
<useCompoundFile>false</useCompoundFile>
<mergeFactor>10</mergeFactor>
Do you thinking lower mergeFactor to 5 and set useCompoundfile to true would
help? I'll try it out on Monday.
Thanks again!
-----Original Message-----
From: Don Werve [mailto:d...@madwombat.com]
Sent: Thursday, June 24, 2010 9:09 PM
To: solr-user@lucene.apache.org
Subject: Re: solr indexing takes a long time and is not reponsive to abort
command
2010/6/25 Ya-Wen Hsu <y...@eline.com>
> This situation doesn't happen consistently. When we only ran the
> problematic core, the indexing took significant longer than usual(4hrs -> 11
> hrs). It ran successful in the end. When we ran indexing for all cores at
> the same time, the problematic core never finished indexing such that we
> have to kill the process. This happened twice already. I'm running it
> parallel again to see if the problem still persists.
>
Off the top of my head:
Have you accidentally opened this core multiple times within the same JVM?
I had the same thing happen to me when I was testing out a Solr interface I
had written under JRuby; that was loads of fun to track down...
How physically large is the core ('du -sh' if you're on Unix), and how many
files does the index contain? I've run into issues where frequent updates
created a lot of index files, and which slowed down all core access.
If you've got a lot of index files, has the problem core been optimized?
