You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@omid.apache.org by "Ohad Shacham (JIRA)" <ji...@apache.org> on 2019/03/20 10:21:00 UTC
[jira] [Resolved] (OMID-125) TSO panic in lease manager causes
deadlock
[ https://issues.apache.org/jira/browse/OMID-125?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Ohad Shacham resolved OMID-125.
-------------------------------
Resolution: Fixed
> TSO panic in lease manager causes deadlock
> ------------------------------------------
>
> Key: OMID-125
> URL: https://issues.apache.org/jira/browse/OMID-125
> Project: Apache Omid
> Issue Type: Sub-task
> Affects Versions: 1.0.1
> Reporter: Philippe Laflamme
> Assignee: Yonatan Gottesman
> Priority: Major
> Fix For: 1.0.1
>
> Attachments: OMID-125.patch
>
>
> When the TSO loses its lease, it panics and invokes the panicker. The default SystemExitPanicker calls System.exit synchronously which, in turn, calls the shutdown hooks synchronously.
> This causes the shutdown hooks to run within the Lease management scheduler's lock which is also required to run shutdown.
> To replicate the issue:
> * start the tso server in HA
> * from a ZK shell, update the tso-lease node:
> ** get /omid/tso-lease
> ** set /omid/tso-lease <the value from get>
> * the tso server will panic but fail to shutdown properly
> A patch is provided here https://github.com/apache/incubator-omid/pull/52
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)