[
https://issues.apache.org/jira/browse/SOLR-4816?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13669328#comment-13669328
]
Mark Miller commented on SOLR-4816:
-----------------------------------
bq. The problem with CloudSolrServer back compat is having the same response.
Not if that response is enabled with a setter? Otherwise you can do the same
thing as now.
bq. If we do a single thread we can just throw an exception and be done with
it. So that would be backwards compatible.
Right, that's my point - you could have a single thread by default - it would
be back compat - then if you flip a switch, you get a thread per server an N
sized thread pool or whatever you are currently doing for this.
bq. But if we go this route I'd like to have an update method with the full
performance enhancements as well.
You shouldn't need this method exposed if options enable the multi threading.
bq. maximizing Solr Cloud indexing performance.
I think until you are using more than a single thread per server, you will be
very very far from easily maxing performance - which is why the other issues
interest me a lot more when considering a concurrency improvement.
> ConcurrentUpdateCloudSolrServer
> -------------------------------
>
> Key: SOLR-4816
> URL: https://issues.apache.org/jira/browse/SOLR-4816
> Project: Solr
> Issue Type: Improvement
> Components: SolrCloud
> Affects Versions: 4.3
> Reporter: Joel Bernstein
> Priority: Minor
> Attachments: SOLR-4816.patch, SOLR-4816.patch, SOLR-4816.patch,
> SOLR-4816.patch, SOLR-4816.patch, SOLR-4816.patch, SOLR-4816.patch,
> SOLR-4816.patch, SOLR-4816.patch, SOLR-4816.patch, SOLR-4816.patch,
> SOLR-4816.patch, SOLR-4816.patch, SOLR-4816.patch, SOLR-4816.patch,
> SOLR-4816.patch, SOLR-4816.patch, SOLR-4816.patch, SOLR-4816.patch,
> SOLR-4816.patch, SOLR-4816-sriesenberg.patch
>
>
> This issue adds a new Solr Cloud client called the
> ConcurrentUpdateCloudSolrServer. This Solr Cloud client implements document
> routing in the client so that document routing overhead is eliminated on the
> Solr servers. Documents are batched up for each shard and then each batch is
> sent in it's own thread.
> With this client, Solr Cloud indexing throughput should scale linearly with
> cluster size.
> This client also has robust failover built-in because the actual requests are
> made using the LBHttpSolrServer. The list of urls used for the request to
> each shard begins with the leader and is followed by that shard's replicas.
> So the leader will be tried first and if it fails it will try the replicas.
> Sample usage:
> ConcurrentUpdateCloudServer client = new
> ConcurrentUpdateCloudSolrServer(zkHostAddress);
> UpdateRequest request = new UpdateRequest();
> SolrInputDocument doc = new SolrInputDocument();
> doc.addField("id", 2);
> doc.addField("manu","BMW");
> request.add(doc);
> NamedList response = client.request(request);
> NamedList exceptions = response.get("exceptions"); // contains any exceptions
> from the shards
> NamedList responses = response.get("responses"); // contains the responses
> from shards without exception.
>
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
