[
https://issues.apache.org/jira/browse/SOLR-6832?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14337817#comment-14337817
]
Otis Gospodnetic commented on SOLR-6832:
----------------------------------------
Hmmmm.... didn't examine the patch or tried this functionality, but based on
your description... here are some comments.
bq. This helps only where there is over-sharding.
That in itself should be avoided whenever possible in my experience. Overhead
around memory and communication during querying. Could be related to your
deadlocks. Or maybe you do a ton more writes so distributing writes across all
nodes is worth the query-time overhead of over-sharding?
bq. Since all queries were send to other nodes, we were getting hit with
distributed deadlocks more often when one or more nodes were slow/overloaded.
Hmmmm... if that is truly happening, then isn't that a separate issue to be
fixed?
bq. So this patch is a slight optimization and a reduction of likelihood of
getting bogged down by other slow nodes when the parent query node has the core.
But VERY slight, right? (hence my Q about whether you've quantified the
improvement from this patch)
Intuitively, querying the local data makes sense - why would one not do that if
the data is right there. I just wonder how much you really benefit if you are
saving just 1 (or very small) number of network calls in request that ends up
dispatching NN requests to NN other nodes in the cluster.
> Queries be served locally rather than being forwarded to another replica
> ------------------------------------------------------------------------
>
> Key: SOLR-6832
> URL: https://issues.apache.org/jira/browse/SOLR-6832
> Project: Solr
> Issue Type: Bug
> Components: SolrCloud
> Affects Versions: 4.10.2
> Reporter: Sachin Goyal
> Assignee: Timothy Potter
> Fix For: Trunk, 5.1
>
> Attachments: SOLR-6832.patch, SOLR-6832.patch, SOLR-6832.patch,
> SOLR-6832.patch
>
>
> Currently, I see that code flow for a query in SolrCloud is as follows:
> For distributed query:
> SolrCore -> SearchHandler.handleRequestBody() -> HttpShardHandler.submit()
> For non-distributed query:
> SolrCore -> SearchHandler.handleRequestBody() -> QueryComponent.process()
> \\
> \\
> \\
> For a distributed query, the request is always sent to all the shards even if
> the originating SolrCore (handling the original distributed query) is a
> replica of one of the shards.
> If the original Solr-Core can check itself before sending http requests for
> any shard, we can probably save some network hopping and gain some
> performance.
> \\
> \\
> We can change SearchHandler.handleRequestBody() or HttpShardHandler.submit()
> to fix this behavior (most likely the former and not the latter).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
