The other caches (filterCache, queryResultCache) only have meaning for
the current searcher opened. Whenever a soft or
hard-commit-with-opensearcher-true, those caches are all reset to 0
and stats start accumulating until the _next_ commit. So depending on
when you look you might see a cache that's just been cleared because a
new searcher has been opened or one that's been in use for a long
time.

However, there is a cumulative section for those two caches. The
cumulative section accumulates the statistics since Solr was started.
That section should be much more stable on a system that's been
running for a while. Of course those all start out at zero when you
start Solr.

Whether the HDFS cache stats follow the same pattern, I don't know.

Best,
Erick
On Mon, Sep 24, 2018 at 3:22 PM Jean-Marc Spaggiari
<jean-m...@spaggiari.org> wrote:
>
> It might be similar to the other cache.. Do you have any pointer the the
> LRUCache documentation? I want to see what kind of metrics it uses and see
> if it's similar or not
> to solr/core/src/java/org/apache/solr/core/HdfsDirectoryFactory.java,,,
>
> Le lun. 24 sept. 2018 à 18:04, Shawn Heisey <apa...@elyograg.org> a écrit :
>
> > On 9/24/2018 3:43 PM, Jean-Marc Spaggiari wrote:
> > > Thanks for taking a look. My indexes are on HDFS. And I configured all
> > the
> > > solr parameters for that. The "shard page" is when I click on a SOLR
> > server
> > > to go to the UI, then in the dropdown on the left I select a shard (a
> > > leader one), then I click on "plugins/stats". And then I open
> > > "HdfsBlockCache".  There from time to time I can see some numbers, but
> > when
> > > I keep refreshing it comes back to 0 always, and I have no clues what is
> > > into the cache, or even if it just work :(
> >
> > Apologies for jumping into a world where I can't offer any help.  I have
> > never used HDFS, so things like this that are related to it are an
> > unknown to me.  I understand most of the other things in Plugins/Stats,
> > but not that one.
> >
> > Thanks,
> > Shawn
> >
> >

Reply via email to