[
https://issues.apache.org/jira/browse/SOLR-16500?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17708192#comment-17708192
]
ASF subversion and git services commented on SOLR-16500:
--------------------------------------------------------
Commit f0d46dede8e5658410d3cb45ba3400dee7d90035 in solr's branch
refs/heads/main from vinayak hegde
[ https://gitbox.apache.org/repos/asf?p=solr.git;h=f0d46dede8e ]
SOLR 16642 : upgrade Solr to use Lucene 9.5.0 (#1360)
* SlowAtomicReader should throw UnsupportedOperationException for all
unimplementable methods, namely the nearest-neighbor ones.
* MMapDirectoryFactory no longer supports "unmap" param but can be done
differently.
* SlowCompositeReaderWrapper can no longer cache Terms.
* ComplexPhrase QParser: can't highlight when hl.usePhraseHighlighter=false due
to UnifiedHighlighter change.
* revert SOLR-16500; Java 19 is supported already by Lucene
* Possible perf regression for timeAllowed param; to be addressed separately
---------
Co-authored-by: David Smiley <[email protected]>
> Enable JDK 19 preview options for MMapDirectory V2
> --------------------------------------------------
>
> Key: SOLR-16500
> URL: https://issues.apache.org/jira/browse/SOLR-16500
> Project: Solr
> Issue Type: Task
> Components: scripts and tools
> Reporter: Christine Poerschke
> Assignee: Kevin Risden
> Priority: Major
> Fix For: main (10.0), 9.2
>
> Time Spent: 50m
> Remaining Estimate: 0h
>
> [~uschindler] wrote on
> https://issues.apache.org/jira/browse/SOLR-16442?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17620361#comment-17620361
> ticket:
> {quote}
> We should also think of updating the startup scripts to pass
> {{--enable-preview}} to JVM if exactly Java 19 is detected. Elasticsearch and
> Opensearch will / have already done this. Not doing this will print a startup
> warning on Java 19 to Java logging.
> The main reason to pass {{--enable-preview}} to the JVM is to allow
> MMapDirectory V2 to be enabled and memory mapping happening in larger junks
> for less overhead and use of kernel resources (see vm.max-map-count sysctl).
> It also reduces risk of crushing JDK on bad written plugins or bugs with
> reopening IndexSearcher/IndexReader.
> See https://github.com/apache/lucene/pull/912 for more details.
> {quote}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
