[
https://issues.apache.org/jira/browse/ASTERIXDB-2196?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16331250#comment-16331250
]
Chen Luo edited comment on ASTERIXDB-2196 at 1/18/18 9:24 PM:
--------------------------------------------------------------
Located the problem, which is a test case timing issue related to the insertion
and flush thread. However, it also reveals another problem that it's possible
(though very rare) that a primary index can be flushed without flushing
secondary indexes because of sharing op tracker for multiple partitions.
All these problems would disappear once we separate the op tracker for each
partition ([https://asterix-gerrit.ics.uci.edu/#/c/2263/]).
was (Author: luochen01):
Located the problem, which is a timing issue related to the insertion and flush
thread. However, it also reveals another problem that it's possible (though
very rare) that a primary index can be flushed without flushing secondary
indexes because of sharing op tracker for multiple partitions.
All these problems would disappear once we separate the op tracker for each
partition (https://asterix-gerrit.ics.uci.edu/#/c/2263/).
> MultiPartitionLSMIndexTest sporadically fails
> ---------------------------------------------
>
> Key: ASTERIXDB-2196
> URL: https://issues.apache.org/jira/browse/ASTERIXDB-2196
> Project: Apache AsterixDB
> Issue Type: Bug
> Components: STO - Storage, TEST - Test Framework
> Reporter: Ian Maxon
> Assignee: Chen Luo
> Priority: Major
> Attachments: testRecycleWhileFlushIsScheduled.log
>
>
> Sometimes the MultiPartitionLSMIndexTest will fail sporadically, with an
> error like " expected:<[1512792358263,1512792358263]> but
> was:<[1512792358261,1512792358261]> "
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
