[jira] [Commented] (SOLR-9824) Documents indexed in bulk are replicated using too many HTTP requests

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

[jira] [Commented] (SOLR-9824) Documents indexed in bulk are replicated using too many HTTP requests

JIRA jira@apache.org

    [ https://issues.apache.org/jira/browse/SOLR-9824?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16020597#comment-16020597 ]

Tomás Fernández Löbbe commented on SOLR-9824:
---------------------------------------------

I just committed the changes of my patch together with SOLR-10233 commit

> Documents indexed in bulk are replicated using too many HTTP requests
> ---------------------------------------------------------------------
>
>                 Key: SOLR-9824
>                 URL: https://issues.apache.org/jira/browse/SOLR-9824
>             Project: Solr
>          Issue Type: Improvement
>      Security Level: Public(Default Security Level. Issues are Public)
>          Components: SolrCloud
>    Affects Versions: 6.3
>            Reporter: David Smiley
>            Assignee: Mark Miller
>         Attachments: SOLR-9824.patch, SOLR-9824.patch, SOLR-9824.patch, SOLR-9824.patch, SOLR-9824.patch, SOLR-9824.patch, SOLR-9824.patch, SOLR-9824-tflobbe.patch
>
>
> This takes awhile to explain; bear with me. While working on bulk indexing small documents, I looked at the logs of my SolrCloud nodes.  I noticed that shards would see an /update log message every ~6ms which is *way* too much.  These are requests from one shard (that isn't a leader/replica for these docs but the recipient from my client) to the target shard leader (no additional replicas).  One might ask why I'm not sending docs to the right shard in the first place; I have a reason but it's besides the point -- there's a real Solr perf problem here and this probably applies equally to replicationFactor>1 situations too.  I could turn off the logs but that would hide useful stuff, and it's disconcerting to me that so many short-lived HTTP requests are happening, somehow at the bequest of DistributedUpdateProcessor.  After lots of analysis and debugging and hair pulling, I finally figured it out.  
> In SOLR-7333 ([~tpot]) introduced an optimization called {{UpdateRequest.isLastDocInBatch()}} in which ConcurrentUpdateSolrClient will poll with a '0' timeout to the internal queue, so that it can close the connection without it hanging around any longer than needed.  This part makes sense to me.  Currently the only spot that has the smarts to set this flag is {{JavaBinUpdateRequestCodec.unmarshal.readOuterMostDocIterator()}} at the last document.  So if a shard received docs in a javabin stream (but not other formats) one would expect the _last_ document to have this flag.  There's even a test.  Docs without this flag get the default poll time; for javabin it's 25ms.  Okay.
> I _suspect_ that if someone used CloudSolrClient or HttpSolrClient to send javabin data in a batch, the intended efficiencies of SOLR-7333 would apply.  I didn't try. In my case, I'm using ConcurrentUpdateSolrClient (and BTW DistributedUpdateProcessor uses CUSC too).  CUSC uses the RequestWriter (defaulting to javabin) to send each document separately without any leading marker or trailing marker.  For the XML format by comparison, there is a leading and trailing marker (<stream> ... </stream>).  Since there's no outer container for the javabin unmarshalling to detect the last document, it marks _every_ document as {{req.lastDocInBatch()}}!  Ouch!



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]