[
https://issues.apache.org/jira/browse/SOLR-9937?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15805747#comment-15805747
]
Mark Miller commented on SOLR-9937:
-----------------------------------
No, the behavior is not correct, we want a move to provide resiliency for
replication and a move means atomic file stuff we want to ensure is the default
behavior. We need to make sure someone else doesn't break this after it's fixed.
> StandardDirectoryFactory::move never uses more efficient implementation
> -----------------------------------------------------------------------
>
> Key: SOLR-9937
> URL: https://issues.apache.org/jira/browse/SOLR-9937
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Mike Drob
> Assignee: Mark Miller
> Attachments: SOLR-9937.patch
>
>
> {noformat}
> Path path1 = ((FSDirectory)
> baseFromDir).getDirectory().toAbsolutePath();
> Path path2 = ((FSDirectory)
> baseFromDir).getDirectory().toAbsolutePath();
>
> try {
> Files.move(path1.resolve(fileName), path2.resolve(fileName),
> StandardCopyOption.ATOMIC_MOVE);
> } catch (AtomicMoveNotSupportedException e) {
> Files.move(path1.resolve(fileName), path2.resolve(fileName));
> }
> {noformat}
> Because {{path1 == path2}} this code never does anything and move always
> defaults to the less efficient implementation in DirectoryFactory.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
