You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Shalin Shekhar Mangar (JIRA)" <ji...@apache.org> on 2013/10/09 17:02:46 UTC
[jira] [Updated] (SOLR-5324) Make sub shard replica recovery and
shard state switch asynchronous
[ https://issues.apache.org/jira/browse/SOLR-5324?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Shalin Shekhar Mangar updated SOLR-5324:
----------------------------------------
Attachment: SOLR-5324.patch
Changes:
# A new shard state: 'recovery' is added
# After all sub-shard replicas have been created, the sub-shard state is set to 'recovery'. If replication factor is 1 then the sub-shards are set to 'active'. The splitshard API returns at this point.
# The state change events in the overseer are used to track when all replicas of all sub-shards become 'active'. Once that happens, the parent shard is set to inactive and the sub-shards are set to 'active'.
# To facilitate the above, a slice property called 'parent' is introduced which is removed once the slice becomes 'active'.
# If the split is retried then we use the 'deleteshard' api to completely remove the sub-shards before starting the splitting process.
> Make sub shard replica recovery and shard state switch asynchronous
> -------------------------------------------------------------------
>
> Key: SOLR-5324
> URL: https://issues.apache.org/jira/browse/SOLR-5324
> Project: Solr
> Issue Type: Improvement
> Components: SolrCloud
> Reporter: Shalin Shekhar Mangar
> Assignee: Shalin Shekhar Mangar
> Fix For: 5.0, 4.6
>
> Attachments: SOLR-5324.patch
>
>
> Currently the shard split command waits for all replicas of all sub shards to recover and then switches the state of parent to inactive and sub-shards to active.
> The problem is that shard split (ab)uses the CoreAdmin WaitForState action to ask the sub shard leader to wait until the replica states are active. This action is prone to timeout.
> We should make the shard state switching asynchronous. Once all replicas of all sub-shards are 'active', the shard states should be switched automatically.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org