accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "John Vines (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-2084) Potential deadlock with namespace reservations in clone table fate operation
Date Mon, 03 Feb 2014 23:55:15 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-2084?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13890140#comment-13890140
] 

John Vines commented on ACCUMULO-2084:
--------------------------------------

Figured it out, needs to lock them in order to ensure they still exist/dont' get deleted out
from underneath the table while the operation is happening

> Potential deadlock with namespace reservations in clone table fate operation
> ----------------------------------------------------------------------------
>
>                 Key: ACCUMULO-2084
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-2084
>             Project: Accumulo
>          Issue Type: Sub-task
>          Components: client, master, tserver
>            Reporter: Christopher Tubbs
>            Assignee: Christopher Tubbs
>             Fix For: 1.6.0
>
>
> Clone table needs to reserve both the source and destination namespace, (as well as the
source tableId) and it needs to do so in a predictable order, not attempting to grab the second
until it gets the first... otherwise, deadlocks could occur. To ensure ordering, we need to
make sure clone table (and other operations) reserve everything up front, reserving the namespaces
first, and the tables second. The namespaces should be sorted before reserving, to ensure
overlaps with other fate operations do not cause deadlocks.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

Mime
View raw message