[
https://issues.apache.org/jira/browse/SOLR-11673?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16284256#comment-16284256
]
Mikhail Khludnev commented on SOLR-11673:
-----------------------------------------
quoting
https://issues.apache.org/jira/browse/SOLR-11293?focusedCommentId=16182379&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16182379
bq. In case of masterVersion = zero, tlog replica won't do commit, just open
new searcher.
Here is what the parameter {{skipCommitOnMasterVersionZero}} was introduced
for.
But the thing is that it always makes sense (all tests are passed). The points
are:
- there is a race condition between initial slave commit and the first
meaningful master commit, the later might just be ignored.
- I don't see a reason behind [that slave
commit|https://github.com/apache/lucene-solr/blame/master/solr/core/src/java/org/apache/solr/handler/IndexFetcher.java#L458]
at all. Could it happen that it was necessary some time ago and now it just
breaks the test(s) rarely?
Thanks for responding, [~tomasflobbe]
> ReplicationHandler race-condition between deleting slave index and commit in
> master
> -----------------------------------------------------------------------------------
>
> Key: SOLR-11673
> URL: https://issues.apache.org/jira/browse/SOLR-11673
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Mikhail Khludnev
> Attachments: SOLR-11673-reproducer.patch,
> SOLR-11673-skipCommitOnMasterVersionZero.patch, SOLR-11673-test-fix.patch,
> doTestIndexAndConfigReplication-consoleText.txt
>
>
> failure in master [described in
> SOLR-6228|https://issues.apache.org/jira/browse/SOLR-6228?focusedCommentId=16266007&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16266007].
> {code}
> 2> NOTE: reproduce with: ant test -Dtestcase=TestReplicationHandler
> -Dtests.method=doTestIndexAndConfigReplication -Dtests.seed=C541E9C9CC845BA5
> -Dtests.slow=true -Dtests.locale=es-BO -Dtests.timezone=Africa/Addis_Ababa
> -Dtests.asserts=true -Dtests.file.encoding=US-ASCII
> [10:13:23.442] ERROR 36.6s |
> TestReplicationHandler.doTestIndexAndConfigReplication <<<
> > Throwable #1: java.lang.IndexOutOfBoundsException: Index: 0, Size: 0
> > at
> __randomizedtesting.SeedInfo.seed([C541E9C9CC845BA5:D109B29CEF83E6BB]:0)
> > at java.util.ArrayList.rangeCheck(ArrayList.java:653)
> > at java.util.ArrayList.get(ArrayList.java:429)
> > at
> org.apache.solr.handler.TestReplicationHandler.doTestIndexAndConfigReplication(TestReplicationHandler.java:561)
> {code}
> Easily reproducible in master by beast.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
