[jira] [Reopened] (AMQ-4897) Race condition in failover transport

["Gary Tully (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/AMQ-4897?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Gary Tully reopened AMQ-4897:
-----------------------------
      Assignee: Gary Tully  (was: Dejan Bosanac)

attempting to lock the reconnectMutex in handle transport failure when a write is blocked results in not being able to timeout the connection. the inactivity monitor blocks leaving a stalled connection.

> Race condition in failover transport
> ------------------------------------
>
>                 Key: AMQ-4897
>                 URL: https://issues.apache.org/jira/browse/AMQ-4897
>             Project: ActiveMQ
>          Issue Type: Bug
>    Affects Versions: 5.9.0
>            Reporter: Dejan Bosanac
>            Assignee: Gary Tully
>             Fix For: 5.9.1, 5.10.0
>
>
> There's a small chance of the race condition when using priority backup with extremely small reconnect delay (0). In that case, the failover transport will get into inconsistent state. The client will stay connected to the "non-priority" broker, but the priority backup will not be created due to this inconsistency.
> The solution is to synchronise handling of connection failure with the reconnect mutex.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)