Hello Otis,
Thank you for the reply.
But, if the high response time is caused by IO and CPU usage... it
shouldn't come back to regular response time after a while? Also,
we're not seeing that much CPU usage.
What if some requests got stuck, and the old searchers are not being
destroyed? Would it account for the high response times?
PS.: Sorry my poor english, I'm not a native speaker :-)
----
Victor Hogemann - Ideais Tecnologia
Tel (RJ) +55 +21 3553-1301 (R217)
Tel (SP) +55 +11 4063-0443 (R217)
Fax +55 +21 3553-1302
Rua da Assembléia, nº 98, 9º andar
Centro | Rio de Janeiro | Brasil
CEP 20011-000
On Dec 29, 2008, at 4:26 PM, Otis Gospodnetic wrote:
Hello Victor,
The high response times could be caused by long and intense warming
of the new searcher(s), which could be causing high IO (it sounds
like all instances share the same index on the same file system) and
high CPU usage (lots of cache items being copied from the old
searcher).
Otis
--
Sematext -- http://sematext.com/ -- Lucene - Solr - Nutch
----- Original Message ----
From: Victor Hogemann <vic...@ideais.com.br>
To: Solr-User <solr-user@lucene.apache.org>
Sent: Monday, December 29, 2008 7:53:44 AM
Subject: High response times after snapshot install
Greetings,
Before anything I'd like to thank the Solr team for such great
piece of
software! Thank you.
Now... back to my problem. I'm experiencing slowdowns on a Solr 1.3
multicore
installation after a snapshot install, response times go up from
5ms to above
5sec, and don't go down unless we restart the servers.
Our index has more than 20 fields, and the slave servers got some
50hits/sec.
We're running it on Weblogic, with 6 Weblogic Nodes/Solr instances
per machine,
sharing the same index on the filesystem.
Any insight on what is happening?
Thanks.
----
Victor Hogemann - Ideais Tecnologia
Tel (RJ) +55 +21 3553-1301 (R217)
Tel (SP) +55 +11 4063-0443 (R217)
Fax +55 +21 3553-1302
Rua da Assembléia, nº 98, 9º andar
Centro | Rio de Janeiro | Brasil
CEP 20011-000