You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Hoss Man (Created) (JIRA)" <ji...@apache.org> on 2011/12/02 22:13:40 UTC
[jira] [Created] (SOLR-2944) Anecdotal evidence that
useFilterForSortedQuery has gotten much slower in 3.5 compared to 1.4
Anecdotal evidence that useFilterForSortedQuery has gotten much slower in 3.5 compared to 1.4
---------------------------------------------------------------------------------------------
Key: SOLR-2944
URL: https://issues.apache.org/jira/browse/SOLR-2944
Project: Solr
Issue Type: Bug
Reporter: Hoss Man
Pawel Rog reported on the user list that he was seeing significantly slower search performance and much higher CPU usage using Solr 3.5 compared to Solr 1.4 when using the same configs he had always used in Solr 1.4 -- even though he had a new index built with 3.5
https://mail-archives.apache.org/mod_mbox/lucene-solr-user/201111.mbox/%3CCAF9ZkbPHo0Y-PvSkm4dkaNc7aArGacfSSHNQBL+nOOt1aHxLwA@mail.gmail.com%3E
Based on stacktraces from threaddumps the "useFilterForSortedQuery" was determined to be a contributing factor, and disabling it improved performance (exact comparison not provided).
Would be nice to get some more details and a reproducible comparison (SOLR-2646 where are you???) and evluate why useFilterForSortedQuery would be slower in 3.5 then in 1.4 with the same data.
Should consider actively removing the useFilterForSortedQuery=true option if there are no longer any usecases where it's better then useFilterForSortedQuery=false
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org
[jira] [Commented] (SOLR-2944) Anecdotal evidence that
useFilterForSortedQuery has gotten much slower in 3.5 compared to 1.4
Posted by "Yonik Seeley (Commented) (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/SOLR-2944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13165820#comment-13165820 ]
Yonik Seeley commented on SOLR-2944:
------------------------------------
I've done some ad-hoc testing using the useFilterForSortedQuery option on 1.4.1 and 3.5.
3.5 has turned out faster in every case I tested (between 1.3 and 2.1 times faster).
The indexes built weren't quite equivalent because of the merge policies.
The 1.4 index had 6 segments, and the 3.5 index had 17 segments.
> Anecdotal evidence that useFilterForSortedQuery has gotten much slower in 3.5 compared to 1.4
> ---------------------------------------------------------------------------------------------
>
> Key: SOLR-2944
> URL: https://issues.apache.org/jira/browse/SOLR-2944
> Project: Solr
> Issue Type: Bug
> Reporter: Hoss Man
>
> Pawel Rog reported on the user list that he was seeing significantly slower search performance and much higher CPU usage using Solr 3.5 compared to Solr 1.4 when using the same configs he had always used in Solr 1.4 -- even though he had a new index built with 3.5
> https://mail-archives.apache.org/mod_mbox/lucene-solr-user/201111.mbox/%3CCAF9ZkbPHo0Y-PvSkm4dkaNc7aArGacfSSHNQBL+nOOt1aHxLwA@mail.gmail.com%3E
> Based on stacktraces from threaddumps the "useFilterForSortedQuery" was determined to be a contributing factor, and disabling it improved performance (exact comparison not provided).
> Would be nice to get some more details and a reproducible comparison (SOLR-2646 where are you???) and evluate why useFilterForSortedQuery would be slower in 3.5 then in 1.4 with the same data.
> Should consider actively removing the useFilterForSortedQuery=true option if there are no longer any usecases where it's better then useFilterForSortedQuery=false
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org
[jira] [Commented] (SOLR-2944) Anecdotal evidence that
useFilterForSortedQuery has gotten much slower in 3.5 compared to 1.4
Posted by "Hoss Man (Commented) (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/SOLR-2944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13161926#comment-13161926 ]
Hoss Man commented on SOLR-2944:
--------------------------------
Uwe: totally possible, but he was upgrading from Solr 1.4, so there are lots of changes on the table as possible culprits.
> Anecdotal evidence that useFilterForSortedQuery has gotten much slower in 3.5 compared to 1.4
> ---------------------------------------------------------------------------------------------
>
> Key: SOLR-2944
> URL: https://issues.apache.org/jira/browse/SOLR-2944
> Project: Solr
> Issue Type: Bug
> Reporter: Hoss Man
>
> Pawel Rog reported on the user list that he was seeing significantly slower search performance and much higher CPU usage using Solr 3.5 compared to Solr 1.4 when using the same configs he had always used in Solr 1.4 -- even though he had a new index built with 3.5
> https://mail-archives.apache.org/mod_mbox/lucene-solr-user/201111.mbox/%3CCAF9ZkbPHo0Y-PvSkm4dkaNc7aArGacfSSHNQBL+nOOt1aHxLwA@mail.gmail.com%3E
> Based on stacktraces from threaddumps the "useFilterForSortedQuery" was determined to be a contributing factor, and disabling it improved performance (exact comparison not provided).
> Would be nice to get some more details and a reproducible comparison (SOLR-2646 where are you???) and evluate why useFilterForSortedQuery would be slower in 3.5 then in 1.4 with the same data.
> Should consider actively removing the useFilterForSortedQuery=true option if there are no longer any usecases where it's better then useFilterForSortedQuery=false
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org
[jira] [Commented] (SOLR-2944) Anecdotal evidence that
useFilterForSortedQuery has gotten much slower in 3.5 compared to 1.4
Posted by "Uwe Schindler (Commented) (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/SOLR-2944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13161916#comment-13161916 ]
Uwe Schindler commented on SOLR-2944:
-------------------------------------
I have the feeling, SOLR-2429 could cause this. There were no changes in Lucene 3.5 except a refactoring in searchWithFilter (moved to FilteredQuery), but thats not even used by Solr (Solr has its own filter logic not using any line of Lucene code most of the time).
> Anecdotal evidence that useFilterForSortedQuery has gotten much slower in 3.5 compared to 1.4
> ---------------------------------------------------------------------------------------------
>
> Key: SOLR-2944
> URL: https://issues.apache.org/jira/browse/SOLR-2944
> Project: Solr
> Issue Type: Bug
> Reporter: Hoss Man
>
> Pawel Rog reported on the user list that he was seeing significantly slower search performance and much higher CPU usage using Solr 3.5 compared to Solr 1.4 when using the same configs he had always used in Solr 1.4 -- even though he had a new index built with 3.5
> https://mail-archives.apache.org/mod_mbox/lucene-solr-user/201111.mbox/%3CCAF9ZkbPHo0Y-PvSkm4dkaNc7aArGacfSSHNQBL+nOOt1aHxLwA@mail.gmail.com%3E
> Based on stacktraces from threaddumps the "useFilterForSortedQuery" was determined to be a contributing factor, and disabling it improved performance (exact comparison not provided).
> Would be nice to get some more details and a reproducible comparison (SOLR-2646 where are you???) and evluate why useFilterForSortedQuery would be slower in 3.5 then in 1.4 with the same data.
> Should consider actively removing the useFilterForSortedQuery=true option if there are no longer any usecases where it's better then useFilterForSortedQuery=false
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org
[jira] [Commented] (SOLR-2944) Anecdotal evidence that
useFilterForSortedQuery has gotten much slower in 3.5 compared to 1.4
Posted by "Uwe Schindler (Commented) (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/SOLR-2944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13161933#comment-13161933 ]
Uwe Schindler commented on SOLR-2944:
-------------------------------------
Maybe we should ask him to also check Solr 3.3 (the above change was introduced in 3.4). I was just going through the logs of SolrIndexSearcher.
> Anecdotal evidence that useFilterForSortedQuery has gotten much slower in 3.5 compared to 1.4
> ---------------------------------------------------------------------------------------------
>
> Key: SOLR-2944
> URL: https://issues.apache.org/jira/browse/SOLR-2944
> Project: Solr
> Issue Type: Bug
> Reporter: Hoss Man
>
> Pawel Rog reported on the user list that he was seeing significantly slower search performance and much higher CPU usage using Solr 3.5 compared to Solr 1.4 when using the same configs he had always used in Solr 1.4 -- even though he had a new index built with 3.5
> https://mail-archives.apache.org/mod_mbox/lucene-solr-user/201111.mbox/%3CCAF9ZkbPHo0Y-PvSkm4dkaNc7aArGacfSSHNQBL+nOOt1aHxLwA@mail.gmail.com%3E
> Based on stacktraces from threaddumps the "useFilterForSortedQuery" was determined to be a contributing factor, and disabling it improved performance (exact comparison not provided).
> Would be nice to get some more details and a reproducible comparison (SOLR-2646 where are you???) and evluate why useFilterForSortedQuery would be slower in 3.5 then in 1.4 with the same data.
> Should consider actively removing the useFilterForSortedQuery=true option if there are no longer any usecases where it's better then useFilterForSortedQuery=false
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org