[jira] [Commented] (SOLR-5501) Ability to work with cold replicas

["Manuel Lenormand (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SOLR-5501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13909726#comment-13909726 ] 

Manuel Lenormand commented on SOLR-5501:
----------------------------------------

clusterstate would look like
{code:javascript}
{"collection1":{
    "shards":{
      "shard1":{
        "range":"00000000-ffffffff",
        "state":"active",
        "replicas":{
          "core_node1":{
            "state":"active",
            "roles":"hot_replica",
            "base_url":"http://127.0.0.1:8983/solr",
            "core":"collection1",
            "node_name":"127.0.0.1:8983_solr"},
          "core_node2":{
            "state":"active",
            "roles":"cold_replica",
            "base_url":"http://10.0.0.4:8983/solr",
            "core":"collection1",
            "node_name":"10.0.0.4:8983_solr",
            "leader":"true"}}}}}
{code}

and the API would be:
solr/admin/collections?action=SETCOLDREPLICA&collection=collection1&shard=shard1&replica=core_node1
solr/admin/collections?action=SETHOTREPLICA&collection=collection1&shard=shard1&replica=core_node1

> Ability to work with cold replicas
> ----------------------------------
>
>                 Key: SOLR-5501
>                 URL: https://issues.apache.org/jira/browse/SOLR-5501
>             Project: Solr
>          Issue Type: Improvement
>          Components: SolrCloud
>    Affects Versions: 4.5.1
>            Reporter: Manuel Lenormand
>              Labels: performance
>             Fix For: 4.7
>
>
> Following this conversation from the mailing list:
> http://lucene.472066.n3.nabble.com/Proposal-for-new-feature-cold-replicas-brainstorming-td4097501.html
> Should give the ability to use replicas mainly as backup cores and not for handling high qps rate. 
> This way you would avoid using the caching ressources (solr and OS) used when routing a query to a replica. 
> With many replicas it's harder hitting the solr cache (same query may hit another replica) and having many replicas on the same instance would cause a useless competition on the OS memory for caching.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org