[ 
https://issues.apache.org/jira/browse/HBASE-4610?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13164835#comment-13164835
 ] 

Jonathan Hsieh commented on HBASE-4610:
---------------------------------------

I had started doing this also -- are you sure you want to keep the 'if (count 
== oldcount && count > 0) break' line?  It was removed on the 0.90 version.

{code}
+    long slept = 0;
     for (int oldcount = countOfRegionServers(); !this.master.isStopped();) {
       Thread.sleep(interval);
+      slept += interval;
       count = countOfRegionServers();
       if (count == oldcount && count > 0) break;
 
       String msg;
+      if (count == oldcount && count >= minToStart && slept >= timeout) {
+        LOG.info("Finished waiting for regionserver count to settle; " +
+            "count=" + count + ", sleptFor=" + slept);
+        break;
{code}

Before and after test, TestMasterFailover seemed flaky for me on the 0.92 
branch.  

Is the plan for this 0.92.0 or 0.92.1?
                
> Port HBASE-3380 (Master failover can split logs of live servers) to 92/trunk 
> (definitely bring in config params, decide if we need to do more to fix the 
> bug)
> -------------------------------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: HBASE-4610
>                 URL: https://issues.apache.org/jira/browse/HBASE-4610
>             Project: HBase
>          Issue Type: Bug
>          Components: master
>    Affects Versions: 0.92.0, 0.94.0
>            Reporter: Jonathan Gray
>            Assignee: Jonathan Gray
>             Fix For: 0.92.1
>
>         Attachments: 4610.txt
>
>
> Over in HBASE-3380 we were having some TestMasterFailover flakiness.  We 
> added some more config parameters to better control the master startup loop 
> where it waits for RS to heartbeat in.  We had thought at the time that 92 
> would have a different solution but it is still relying on heartbeats to 
> learn about RSs.
> For now, we should definitely bring these config params into 92/trunk.  
> Otherwise this is an incompatible regression and adding these will also make 
> things like what was just reported over in HBASE-4603 trivial to fix in an 
> optimal way.

--
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