You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-dev@hadoop.apache.org by "Jim Kellerman (JIRA)" <ji...@apache.org> on 2007/11/08 18:47:50 UTC
[jira] Created: (HADOOP-2173) [hbase] When the master times out a
region servers lease, the region server may not restart
[hbase] When the master times out a region servers lease, the region server may not restart
-------------------------------------------------------------------------------------------
Key: HADOOP-2173
URL: https://issues.apache.org/jira/browse/HADOOP-2173
Project: Hadoop
Issue Type: Bug
Components: contrib/hbase
Reporter: Jim Kellerman
Hadoop-Nightly 297 failed because:
* The region server's lease expired (Why? was the heartbeat thread starved?)
* The region server gets a call startup message
* The master splits the region server's log and deletes it.
I think that when the region server called log.closeAndDelete(), it got an exception (because the file no longer existed) at that point it said "error restarting server" and quit. From there on the master is just looping because there is no region server to talk to
We should probably just log an error for log.closeAndDelete() and proceed with region server restart.
Also for that test, we should probably increase the lease timeout and make the lease timeout check happen less frequently accordingly
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Resolved: (HADOOP-2173) [hbase] When the master times out a
region servers lease, the region server may not restart
Posted by "Jim Kellerman (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-2173?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jim Kellerman resolved HADOOP-2173.
-----------------------------------
Resolution: Fixed
Fix Version/s: 0.16.0
> [hbase] When the master times out a region servers lease, the region server may not restart
> -------------------------------------------------------------------------------------------
>
> Key: HADOOP-2173
> URL: https://issues.apache.org/jira/browse/HADOOP-2173
> Project: Hadoop
> Issue Type: Bug
> Components: contrib/hbase
> Reporter: Jim Kellerman
> Assignee: Jim Kellerman
> Fix For: 0.16.0
>
>
> Hadoop-Nightly 297 failed because:
> * The region server's lease expired (Why? was the heartbeat thread starved?)
> * The region server gets a call startup message
> * The master splits the region server's log and deletes it.
> I think that when the region server called log.closeAndDelete(), it got an exception (because the file no longer existed) at that point it said "error restarting server" and quit. From there on the master is just looping because there is no region server to talk to
> We should probably just log an error for log.closeAndDelete() and proceed with region server restart.
> Also for that test, we should probably increase the lease timeout and make the lease timeout check happen less frequently accordingly
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HADOOP-2173) [hbase] When the master times out a
region servers lease, the region server may not restart
Posted by "Hudson (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-2173?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12545819 ]
Hudson commented on HADOOP-2173:
--------------------------------
Integrated in Hadoop-Nightly #315 (See [http://lucene.zones.apache.org:8080/hudson/job/Hadoop-Nightly/315/])
> [hbase] When the master times out a region servers lease, the region server may not restart
> -------------------------------------------------------------------------------------------
>
> Key: HADOOP-2173
> URL: https://issues.apache.org/jira/browse/HADOOP-2173
> Project: Hadoop
> Issue Type: Bug
> Components: contrib/hbase
> Reporter: Jim Kellerman
> Assignee: Jim Kellerman
>
> Hadoop-Nightly 297 failed because:
> * The region server's lease expired (Why? was the heartbeat thread starved?)
> * The region server gets a call startup message
> * The master splits the region server's log and deletes it.
> I think that when the region server called log.closeAndDelete(), it got an exception (because the file no longer existed) at that point it said "error restarting server" and quit. From there on the master is just looping because there is no region server to talk to
> We should probably just log an error for log.closeAndDelete() and proceed with region server restart.
> Also for that test, we should probably increase the lease timeout and make the lease timeout check happen less frequently accordingly
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HADOOP-2173) [hbase] When the master times out a
region servers lease, the region server may not restart
Posted by "Jim Kellerman (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-2173?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12545737 ]
Jim Kellerman commented on HADOOP-2173:
---------------------------------------
This should be fixed with the commit for HADOOP-2276. Leaving open in case other other circumstances also exhibit this bug.
> [hbase] When the master times out a region servers lease, the region server may not restart
> -------------------------------------------------------------------------------------------
>
> Key: HADOOP-2173
> URL: https://issues.apache.org/jira/browse/HADOOP-2173
> Project: Hadoop
> Issue Type: Bug
> Components: contrib/hbase
> Reporter: Jim Kellerman
> Assignee: Jim Kellerman
>
> Hadoop-Nightly 297 failed because:
> * The region server's lease expired (Why? was the heartbeat thread starved?)
> * The region server gets a call startup message
> * The master splits the region server's log and deletes it.
> I think that when the region server called log.closeAndDelete(), it got an exception (because the file no longer existed) at that point it said "error restarting server" and quit. From there on the master is just looping because there is no region server to talk to
> We should probably just log an error for log.closeAndDelete() and proceed with region server restart.
> Also for that test, we should probably increase the lease timeout and make the lease timeout check happen less frequently accordingly
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Assigned: (HADOOP-2173) [hbase] When the master times out a
region servers lease, the region server may not restart
Posted by "Jim Kellerman (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-2173?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jim Kellerman reassigned HADOOP-2173:
-------------------------------------
Assignee: Jim Kellerman
> [hbase] When the master times out a region servers lease, the region server may not restart
> -------------------------------------------------------------------------------------------
>
> Key: HADOOP-2173
> URL: https://issues.apache.org/jira/browse/HADOOP-2173
> Project: Hadoop
> Issue Type: Bug
> Components: contrib/hbase
> Reporter: Jim Kellerman
> Assignee: Jim Kellerman
>
> Hadoop-Nightly 297 failed because:
> * The region server's lease expired (Why? was the heartbeat thread starved?)
> * The region server gets a call startup message
> * The master splits the region server's log and deletes it.
> I think that when the region server called log.closeAndDelete(), it got an exception (because the file no longer existed) at that point it said "error restarting server" and quit. From there on the master is just looping because there is no region server to talk to
> We should probably just log an error for log.closeAndDelete() and proceed with region server restart.
> Also for that test, we should probably increase the lease timeout and make the lease timeout check happen less frequently accordingly
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.