[
https://issues.apache.org/jira/browse/SOLR-4764?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13679245#comment-13679245
]
Michael Garski commented on SOLR-4764:
--------------------------------------
I discovered the same thing, and that there is no re-use of segment caches on a
read-only replication slave as well. I opened SOLR-4909 to handle the slave
replication case. I'm not sure that changing the IndexReaderFactory's method
will work with a read-only slave, but adding another method for NRT/replication
masters that take a writer might work.
> When using NRT, just init the reader from IndexWriter
> -----------------------------------------------------
>
> Key: SOLR-4764
> URL: https://issues.apache.org/jira/browse/SOLR-4764
> Project: Solr
> Issue Type: Improvement
> Reporter: Robert Muir
>
> Spinoff from SOLR-4761
> Solr first opens a DirectoryReader from the directory, then later will pass
> this to IW openIfChanged.
> I noticed this when i was confused that mergedsegmentwarmer doesn't appear to
> work at first until after you've reopened...
> I'm not totally sure what the current behavior causes (does IW's pool reuse
> segments from this passed-in "external" reader, or is this causing some
> horrible doubling-up/inefficient stuff etc?). To some extent i think we
> should change it even if its actually performant: I think its confusing.
> I think ideally we'd change IndexReaderFactory's method to take writer
> instead of directory so that custom DirectoryReaders can still work.
--
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]
