Sorry for not following up sooner- been a busy last couple weeks.
We do see a significant instanity count - could this be due to
updating indexes from the dev Solr build? E.g. on one server I see
<stat name="insanity_count" >
61
</stat>
and entries like:
<stat name="insanity#0" >
SUBREADER: Found caches for decendents of
org.apache.lucene.index.readonlydirectoryrea...@2b8d6cbf+created
'org.apache.lucene.index.readonlydirectoryrea...@2b8d6cbf'=>'created',class
org.apache.lucene.search.FieldCache$StringIndex,null=>org.apache.lucene.search.FieldCache$StringIndex#2002656056
(size =~ 74.4 KB)
'org.apache.lucene.store.niofsdirectory$niofsindexin...@47adeb94'=>'created',class
org.apache.lucene.search.FieldCache$StringIndex,null=>org.apache.lucene.search.FieldCache$StringIndex#1099177573
(size =~ 74.4 KB)
</stat>
<stat name="insanity#1" >
SUBREADER: Found caches for decendents of
org.apache.lucene.index.readonlydirectoryrea...@d0340a9+created
'org.apache.lucene.index.readonlydirectoryrea...@d0340a9'=>'created',class
org.apache.lucene.search.FieldCache$StringIndex,null=>org.apache.lucene.search.FieldCache$StringIndex#868132357
(size =~ 831.2 KB)
'org.apache.lucene.store.niofsdirectory$niofsindexin...@78802615'=>'created',class
org.apache.lucene.search.FieldCache$StringIndex,null=>org.apache.lucene.search.FieldCache$StringIndex#1542727931
(size =~ 831.2 KB)
</stat>
And I think it's higher on the one associated with the screenshot.
using the lucene checkIndex tool does not show any errors.
Most of what we want is returned by the Luke handler, except for the
pending adds and deletes and the index size. I can hack around this
by creating a greatly reduced stats.jsp, but I'd also liek to
understand what we are experiencing.
-Peter
On Fri, Jan 8, 2010 at 1:38 PM, Mark Miller <markrmil...@gmail.com> wrote:
> Yonik Seeley wrote:
>> On Fri, Jan 8, 2010 at 1:03 PM, Mark Miller <markrmil...@gmail.com> wrote:
>>
>>> It should be fixed in trunk, but that was after 1.4. Currently, it
>>> should only do it if it sees insanity - which there shouldn't be any
>>> with stock Solr.
>>>
>>
>> http://svn.apache.org/viewvc/lucene/solr/tags/release-1.4.0/src/java/org/apache/solr/search/SolrFieldCacheMBean.java
>> http://svn.apache.org/viewvc?view=revision&revision=826788
>> Seems like it's there? Or was it a different commit?
>>
>> Perhaps there is just real instanity... which may be unavoidable at
>> this point since not everything in solr is done per-segment yet.
>>
>> -Yonik
>> http://www.lucidimagination.com
>>
>
> Your right - when looking at the Solr release date, I quickly took the
> 10 as October - but it was 11/10, so it is in 1.4.
>
> So people seeing this should also being seeing an insanity count over one.
>
> I'd think that would be rarer than one this sounds like though ... whats
> left that could cause insanity?
>
> We should prob switch to never calculating the size unless an explicit
> param is pass to the stats page.
>
>
> --
> - Mark
>
> http://www.lucidimagination.com
>
>
>
>
--
Peter M. Wolanin, Ph.D.
Momentum Specialist, Acquia. Inc.
peter.wola...@acquia.com
