Per-segment faceting and soft commits

[Jonathan Acheson <gs...@gmail.com>]

Using solr 4.0, I have 2 fields that I'm faceting on using facet.method=fcs.
After the first facet query I can see the two fields in the fieldValueCache.
If I
then issue a soft commit at /update/json?softCommit=true and again look at
the
fieldValueCache the two entries are now gone.

I was thinking that since the commit did not create a new segement the
caches would be reused but the query time and absence of the cache values
indicate
everything was flushed.

Am I incorrectly using this feature and/or incorrectly assuming the
fieldValueCache
should stay populated? Any clarification would be greatly appreciated,
thanks!



--
View this message in context: http://lucene.472066.n3.nabble.com/Per-segment-faceting-and-soft-commits-tp4022538.html
Sent from the Solr - User mailing list archive at Nabble.com.

Re: Per-segment faceting and soft commits

[Mikhail Khludnev <mk...@griddynamics.com>]

Jonathan,

I wonder why softCommit does not create a new segment? I thought it doesn't
call flush/fsync for a new segment. Can you give a reference to your
explanation?
Then, I guess, as far "fieldValueCache" is not a Solr cache but monitoring
wrapper ontop of Lucene per-segment cache, what you see just says that
fieldcache has not been warmed for the last new segment.
Only that segment will create fieldCache on the folowing request. You've
got the trouble only if somewhere you obtain fieldcache for top level
searcher.
 27.11.2012 7:38 пользователь "Jonathan Acheson" <gs...@gmail.com>
написал:

> Using solr 4.0, I have 2 fields that I'm faceting on using
> facet.method=fcs.
> After the first facet query I can see the two fields in the
> fieldValueCache.
> If I
> then issue a soft commit at /update/json?softCommit=true and again look at
> the
> fieldValueCache the two entries are now gone.
>
> I was thinking that since the commit did not create a new segement the
> caches would be reused but the query time and absence of the cache values
> indicate
> everything was flushed.
>
> Am I incorrectly using this feature and/or incorrectly assuming the
> fieldValueCache
> should stay populated? Any clarification would be greatly appreciated,
> thanks!
>
>
>
> --
> View this message in context:
> http://lucene.472066.n3.nabble.com/Per-segment-faceting-and-soft-commits-tp4022538.html
> Sent from the Solr - User mailing list archive at Nabble.com.
>