[ 
https://issues.apache.org/jira/browse/HADOOP-7938?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13188943#comment-13188943
 ] 

Todd Lipcon commented on HADOOP-7938:
-------------------------------------

{code}
+    // Try to fence fromSvc
+    if (fencer != null) {
+      if (!fencer.fence()) {
+        throw new FailoverFailedException("Unable to fence " + fromSvcName);
+      }
     }
{code}

Shouldn't you only fence the old node in the case that you got an exception in 
the {{transitionToStandby}} call? Then, you also wouldn't need to make the user 
specify whether or not to fence - it would automatically fence in the case that 
there was a problem with graceful failover.

----

- If the {{transitionToActive}} call fails, we need to be careful before doing 
a failback. For example, what if it was a timeout, and in fact the new active 
is still in the process of failing over? Then we need to fence "toSvc" before 
going back to "fromSvc"

                
> HA: the FailoverController should optionally fence the active during failover
> -----------------------------------------------------------------------------
>
>                 Key: HADOOP-7938
>                 URL: https://issues.apache.org/jira/browse/HADOOP-7938
>             Project: Hadoop Common
>          Issue Type: Sub-task
>          Components: ha
>    Affects Versions: HA Branch (HDFS-1623)
>            Reporter: Eli Collins
>            Assignee: Eli Collins
>             Fix For: HA Branch (HDFS-1623)
>
>         Attachments: hadoop-7938.txt
>
>
> The FailoverController in HADOOP-7924 needs to be able to fence off the 
> current active in case it fails to transition to standby (or the user 
> requests it for sanity). This is needed even for manual failover (the CLI 
> should use the configured fencing mechanism). The FC needs to access the 
> HDFS-specific implementations HDFS-2179, could add a common fencing interface 
> (or just shell out but we may not always want to do that).

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

        

Reply via email to