[jboss-jira] [JBoss JIRA] (JGRP-1428) UnicastRequest and GroupRequest should mark a target as suspected if the target has already left the cluster at creation time

[Dan Berindei (JIRA)]

     [ https://issues.jboss.org/browse/JGRP-1428?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Dan Berindei reopened JGRP-1428:
--------------------------------



We have reproduced the issue again with JGroups 3.0.6 and 500 threads doing unicast requests at the time of the leave.

The problem could be that RequestCorrelator first runs the suspect check on all requests and only then updates the view reference. Since the ConcurrentHashMap iterator ignores updates, it's possible for our request thread to add the request to the map and check the view membership between the start of the iteration and the view update.
                
> UnicastRequest and GroupRequest should mark a target as suspected if the target has already left the cluster at creation time
> -----------------------------------------------------------------------------------------------------------------------------
>
>                 Key: JGRP-1428
>                 URL: https://issues.jboss.org/browse/JGRP-1428
>             Project: JGroups
>          Issue Type: Bug
>    Affects Versions: 3.0.5, 3.1
>            Reporter: Dan Berindei
>            Assignee: Bela Ban
>            Priority: Critical
>             Fix For: 3.0.6, 3.1
>
>
> UnicastRequest and GroupsRequest both detect if a target has left the cluster and mark its response as "suspected".
> However, if the node has already left the cluster by the time the request is created, its response is not marked as "suspected" and the request never properly finishes.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.jboss.org/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira