[
https://issues.apache.org/jira/browse/LUCENE-8226?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16415456#comment-16415456
]
Robert Muir commented on LUCENE-8226:
-------------------------------------
Maybe the testRandom3 can be pulled out of this into a TestIndexSortingRandom,
and slap the annotation on that.
Because there are also a lot of nice simple unit tests in here that will be no
problem at all: why disable any codec for those? Even most of the other random
tests here are reasonable and use document counts of, e.g. 100.
In general, it seems like only one method is problematic. I don't understand
the justification of a million documents for this test, are there
sorting-specific bugs that it will really find as opposed to a smaller count?
{code}
// pits index time sorting against query time sorting
public void testRandom3() throws Exception {
int numDocs;
if (TEST_NIGHTLY) {
numDocs = atLeast(100000);
} else {
numDocs = atLeast(1000);
}
{code}
> Don't use MemoryCodec for nightly runs of TestIndexSorting
> ----------------------------------------------------------
>
> Key: LUCENE-8226
> URL: https://issues.apache.org/jira/browse/LUCENE-8226
> Project: Lucene - Core
> Issue Type: Task
> Reporter: Alan Woodward
> Priority: Major
>
> Nightly runs of TestIndexSorting fail occasionally with OOM (see
> [https://builds.apache.org/job/Lucene-Solr-NightlyTests-7.x/183/] for a
> recent example, and it's been appearing in Erick's BadApple report too). It
> looks as this is normally due to the combination of a large docset and
> MemoryCodec. We should suppress MemoryCodec for these tests, on nightly runs
> only if possible)
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
