[
https://issues.apache.org/jira/browse/IGNITE-5003?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16721357#comment-16721357
]
Alexey Goncharuk edited comment on IGNITE-5003 at 12/14/18 12:37 PM:
---------------------------------------------------------------------
[~aealeksandrov], agree with Vyacheslav, looks like there is a bug and value
contents is read under a race. {{load(key)}} concern is also valid. Please add
a test with a small data region, eviction policy configured and a concurrent
test which writes and reads several keys, but each thread has it's own set of
keys. Every thread should always read the last written value.
I also added some comments to PR, please take a look.
was (Author: agoncharuk):
[~aealeksandrov], agree with Vyacheslav, looks like there is a bug and value
contents is read under a race. {{load(key)}} concern is also valid. Please add
a test with a small data region, eviction policy configured and a concurrent
test which writes and reads several keys, but each thread has it's own set of
keys. Every thread should always read the last written value.
> Parallel write&evict same key in CacheWriteBehindStore
> ------------------------------------------------------
>
> Key: IGNITE-5003
> URL: https://issues.apache.org/jira/browse/IGNITE-5003
> Project: Ignite
> Issue Type: Improvement
> Affects Versions: 1.9
> Reporter: Alexander Belyak
> Assignee: Andrey Aleksandrov
> Priority: Major
> Fix For: 2.8
>
>
> Now GridCacheWriteBehindStore.updateCache wait for writeLock in StatefulValue
> and, moreover, waitForFlush() if value is in pending (flushing) state.
> We need to remove waiting.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
