cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Blake Eggleston (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (CASSANDRA-13751) Race / ref leak in PendingRepairManager
Date Tue, 08 Aug 2017 17:39:00 GMT

     [ https://issues.apache.org/jira/browse/CASSANDRA-13751?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Blake Eggleston updated CASSANDRA-13751:
----------------------------------------
    Reviewer: Marcus Eriksson
      Status: Patch Available  (was: Open)

[trunk|https://github.com/bdeggleston/cassandra/tree/13751]
[utest|https://circleci.com/gh/bdeggleston/cassandra/76]
[dtest (pending)|https://builds.apache.org/view/A-D/view/Cassandra/job/Cassandra-devbranch-dtest/171/]

> Race / ref leak in PendingRepairManager
> ---------------------------------------
>
>                 Key: CASSANDRA-13751
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-13751
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: Blake Eggleston
>            Assignee: Blake Eggleston
>            Priority: Minor
>             Fix For: 4.0
>
>
> PendingRepairManager#getScanners has an assertion that confirms an sstable is, in fact,
marked as pending repair. Since validation compactions don't use the same concurrency controls
as proper compactions, they can race with promotion/demotion compactions and end up getting
assertion errors when the pending repair id is changed while the scanners are being acquired.
Also, error handling in PendingRepairManager and CompactionStrategyManager leaks refs when
this happens.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@cassandra.apache.org
For additional commands, e-mail: commits-help@cassandra.apache.org


Mime
View raw message