[
https://issues.apache.org/jira/browse/LUCENE-5263?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13789009#comment-13789009
]
Shai Erera commented on LUCENE-5263:
------------------------------------
I ported this test to verify NDV updates aren't lost under such conditions, and
I implemented the TODO about sub-classing CMS. Perhaps you'll want to include
that in your patch as well:
{code}
final MergeScheduler ms = iwc.getMergeScheduler();
if (ms instanceof ConcurrentMergeScheduler) {
final ConcurrentMergeScheduler suppressFakeIOE = new
ConcurrentMergeScheduler() {
@Override
protected void handleMergeException(Throwable exc) {
// suppress FakeIOExceptions
if (!(exc instanceof FakeIOException)) {
super.handleMergeException(exc);
}
}
};
final ConcurrentMergeScheduler cms = (ConcurrentMergeScheduler) ms;
suppressFakeIOE.setMaxMergesAndThreads(cms.getMaxMergeCount(),
cms.getMaxThreadCount());
suppressFakeIOE.setMergeThreadPriority(cms.getMergeThreadPriority());
iwc.setMergeScheduler(suppressFakeIOE);
iwc.setMergeScheduler(suppressFakeIOE);
}
{code}
> Deletes may be silently lost if disk fills up and then frees up
> ---------------------------------------------------------------
>
> Key: LUCENE-5263
> URL: https://issues.apache.org/jira/browse/LUCENE-5263
> Project: Lucene - Core
> Issue Type: Bug
> Components: core/index
> Reporter: Michael McCandless
> Assignee: Michael McCandless
> Fix For: 5.0, 4.6
>
> Attachments: LUCENE-5263.patch
>
>
> This case is tricky to handle, yet I think realistic: disk fills up
> temporarily, causes an exception in writeLiveDocs, and then the app
> keeps using the IW instance.
> Meanwhile disk later frees up again, IW is closed "successfully". In
> certain cases, we can silently lose deletes in this case.
> I had already committed
> TestIndexWriterDeletes.testNoLostDeletesOnDiskFull, and Jenkins seems
> happy with it so far, but when I added fangs to the test (cutover to
> RandomIndexWriter from IndexWriter, allow IOE during getReader, add
> randomness to when exc is thrown, etc.), it uncovered some real/nasty
> bugs:
> * ReaderPool.dropAll was suppressing any exception it hit, because
> {code}if (priorE != null){code} should instead be {code}if (priorE ==
> null){code}
> * After a merge, we have to write deletes before committing the
> segment, because an exception when writing deletes means we need
> to abort the merge
> * Several places that were directly calling deleter.checkpoint must
> also increment the changeCount else on close IW thinks there are
> no changes and doesn't write a new segments file.
> * closeInternal was dropping pooled readers after writing the
> segments file, which would lose deletes still buffered due to a
> previous exc.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
