[
https://issues.apache.org/jira/browse/SOLR-12674?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16585170#comment-16585170
]
Varun Thacker commented on SOLR-12674:
--------------------------------------
After this patch the hash query parser is only used when parallel stream is has
>=2 workers.
This also addresses the problem described on SOLR-12624
> RollupStream should not use the HashQueryParser for 1 worker
> ------------------------------------------------------------
>
> Key: SOLR-12674
> URL: https://issues.apache.org/jira/browse/SOLR-12674
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Varun Thacker
> Priority: Major
> Attachments: SOLR-12674.patch
>
>
> Let's say I have a dataset of a 100M document
> After applying a filter the number of documents found would be 5k so it's
> tiny.
> If I do a search and a rollup stream the query returns in the 200ms range.
> But if by mistake I add the "partitionKeys" param to the search stream the
> hash query parser is invoked which runs on the entire document set and the
> query time spikes up to 7 seconds.
> If we aren't providing a parallel stream we should ignore the partitionKeys
> param
> Sample Query:
> {code:java}
> rollup(search(gettingstarted,q="*:*",fl="id",sort="id
> desc",partitionKeys="id"),over="id"){code}
> Because of the partitionKeys the underlying query formed is:
> {code:java}
> params={q=*:*&distrib=false&fl=id&sort=id+desc&partitionKeys=id&fq={!hash+workers%3D1+worker%3D0}&wt=json&version=2.2}
> hits=2 status=0 QTime=30{code}
> This is a dummy dataset so don't see the the hits and QTime but this query
> certainly doesn't need to add the hash query parser filter clause for
> workers=1
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
