[
https://issues.apache.org/jira/browse/SOLR-5808?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14661878#comment-14661878
]
Guido commented on SOLR-5808:
-----------------------------
Hello, I just ran into this problem as well with Solr version 5.2.1. While you
try to fix this problem, can you please confirm if the Lucene
MultiPassIndexSplitter can actually be a workaround for this problem? If it can
be a workaround, can you please suggest me how to manually do what the 'split
shard' does by using the Lucene MultiPassIndexSplitter?
I suppose that I need to stop ZK and Solr and then launch the split; once it
finishes, restart Zk and Solr. But, for instance, since I am using the
out-of-the-box composite id routing, should I set the 'seq' parameter to
'false'? And do I need to do anything to notify to Solr that the shard has been
splitted in order to properly route the documents that I will add in the future
(maybe by manually modifying the state.json)?
Sorry for it and I understand that this bug fix requires time, but
unfortunately I cannot wait furthermore to split the shard and I am trying to
find a workaround as soon as possible.
Thank you very much
> collections?action=SPLITSHARD running out of heap space due to large segments
> -----------------------------------------------------------------------------
>
> Key: SOLR-5808
> URL: https://issues.apache.org/jira/browse/SOLR-5808
> Project: Solr
> Issue Type: Bug
> Components: update
> Affects Versions: 4.7
> Reporter: Will Butler
> Assignee: Shalin Shekhar Mangar
> Labels: outofmemory, shard, split
>
> This issue is related to [https://issues.apache.org/jira/browse/SOLR-5214].
> Although memory issues due to merging have been resolved, we still run out of
> memory when splitting a shard containing a large segment (created by
> optimizing). The Lucene MultiPassIndexSplitter is able to split the index
> without error.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
