You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "chunhui shen (JIRA)" <ji...@apache.org> on 2012/05/02 07:49:12 UTC
[jira] [Created] (HBASE-5914) Using bulk assign regions in the
process of ServerShutdownHandler
chunhui shen created HBASE-5914:
-----------------------------------
Summary: Using bulk assign regions in the process of ServerShutdownHandler
Key: HBASE-5914
URL: https://issues.apache.org/jira/browse/HBASE-5914
Project: HBase
Issue Type: Improvement
Reporter: chunhui shen
Assignee: chunhui shen
In the process of ServerShutdownHandler, we currently assign regions singly.
In the large cluster, one regionserver always carried many regions, this action is quite slow.
What about using bulk assign regions like cluster start up.
In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "Hadoop QA (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13268890#comment-13268890 ]
Hadoop QA commented on HBASE-5914:
----------------------------------
-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12525380/HBASE-5914v2.patch
against trunk revision .
+1 @author. The patch does not contain any @author tags.
-1 tests included. The patch doesn't appear to include any new or modified tests.
Please justify why no new tests are needed for this patch.
Also please list what manual steps were performed to verify this patch.
+1 hadoop23. The patch compiles against the hadoop 0.23.x profile.
+1 javadoc. The javadoc tool did not generate any warning messages.
+1 javac. The applied patch does not increase the total number of javac compiler warnings.
+1 findbugs. The patch does not introduce any new Findbugs (version 1.3.9) warnings.
+1 release audit. The applied patch does not increase the total number of release audit warnings.
-1 core tests. The patch failed these unit tests:
org.apache.hadoop.hbase.regionserver.TestSplitTransactionOnCluster
org.apache.hadoop.hbase.TestDrainingServer
org.apache.hadoop.hbase.master.TestAssignmentManager
Test results: https://builds.apache.org/job/PreCommit-HBASE-Build/1778//testReport/
Findbugs warnings: https://builds.apache.org/job/PreCommit-HBASE-Build/1778//artifact/trunk/patchprocess/newPatchFindbugsWarnings.html
Console output: https://builds.apache.org/job/PreCommit-HBASE-Build/1778//console
This message is automatically generated.
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "stack (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13271610#comment-13271610 ]
stack commented on HBASE-5914:
------------------------------
+1 on v3. Nice numbers in your test too Chunhui.
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch, HBASE-5914v3.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "ramkrishna.s.vasudevan (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13271600#comment-13271600 ]
ramkrishna.s.vasudevan commented on HBASE-5914:
-----------------------------------------------
I am +1 on this. But did not do any tests.
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch, HBASE-5914v3.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Updated] (HBASE-5914) Bulk assign regions in the process of
ServerShutdownHandler
Posted by "Zhihong Yu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Zhihong Yu updated HBASE-5914:
------------------------------
Status: Patch Available (was: Open)
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "stack (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13270822#comment-13270822 ]
stack commented on HBASE-5914:
------------------------------
Could you call the new method assign overloading existing assigns rather than quickAssign?
Do you know why the above tests failed? Is it related?
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "Zhihong Yu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13267127#comment-13267127 ]
Zhihong Yu commented on HBASE-5914:
-----------------------------------
bq. I think we needn't whether servers is empty, if it is empty, bulkPlan is null
Is that so ?
{code}
+ servers.removeAll(failedPlans.keySet());
{code}
If servers become empty (failedPlans is not empty), we don't need to execute the following loop:
{code}
+ for (Map.Entry<ServerName, List<HRegionInfo>> e : failedPlans.entrySet()) {
{code}
Coming out of the top-most quickAssign() call, how do we know that regions have been assigned ?
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Using bulk assign regions in the
process of ServerShutdownHandler
Posted by "ramkrishna.s.vasudevan (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13266620#comment-13266620 ]
ramkrishna.s.vasudevan commented on HBASE-5914:
-----------------------------------------------
@Chunhui
Idea is good.
But do we need
{code}
boolean waitUntilNoRegionsInTransition(final long timeout)
{code}
such wait mechanisms? Also can we have a configuration like how we have in the master start up to ensure whether to use roundRobin or not.
Because if i have a custom load balancer i will get a call back in this case for round robin. May be we can avoid that if we have the configuration? What do you feel Chunhui?
> Using bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Attachments: HBASE-5914.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Updated] (HBASE-5914) Bulk assign regions in the process of
ServerShutdownHandler
Posted by "chunhui shen (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
chunhui shen updated HBASE-5914:
--------------------------------
Attachment: HBASE-5914v3.patch
Patch v3 change the method name from quickAssign to assign.
All failed tests are passed.
Also, I have done a performance test.
2012-05-09 14:29:45,364 INFO org.apache.hadoop.hbase.master.handler.ServerShutdownHandler:
Recovery time calculation: reassigning 4000 regions, took 14847ms
2012-05-09 14:38:00,695 INFO org.apache.hadoop.hbase.master.handler.ServerShutdownHandler:
Recovery time calculation: reassigning 4001 regions, took 3242ms
With the patch, the took time of reassigning 4000 regions in the process of SSH is reduced from 14s to 3s.
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch, HBASE-5914v3.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "Zhihong Yu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13271615#comment-13271615 ]
Zhihong Yu commented on HBASE-5914:
-----------------------------------
Integrated to trunk.
Thanks for the patch, Chunhui.
Thanks for the review, Stack and Rama.
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch, HBASE-5914v3.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Updated] (HBASE-5914) Using bulk assign regions in the
process of ServerShutdownHandler
Posted by "chunhui shen (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
chunhui shen updated HBASE-5914:
--------------------------------
Attachment: HBASE-5914.patch
In the HBASE-5914.patch,
We create a method AssignmentManager#quickAssign:
It first bulk assign regions with bulk plan to available servers,
If some is failed, remove servers from available servers and retry
> Using bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Attachments: HBASE-5914.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "ramkrishna.s.vasudevan (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13267154#comment-13267154 ]
ramkrishna.s.vasudevan commented on HBASE-5914:
-----------------------------------------------
@Chunhui
Yes, in SSH case retain assignment is not valid. But what if i don't need the roundrobin assignment itself to be called like how we have in master startup? I just meant a similar configuration like the 'hbase.master.startup.retainassign', for eg.,
'hbase.master.servershutdown.roundrobinassign'something like this. Just my thoughts.
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Using bulk assign regions in the
process of ServerShutdownHandler
Posted by "chunhui shen (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13267107#comment-13267107 ]
chunhui shen commented on HBASE-5914:
-------------------------------------
@ram
In current process of ServerShutdownHandler, we don't wait until no regions in transition, and we could keep it.
bq.can we have a configuration like how we have in the master start up to ensure whether to use roundRobin or not.
When master startup, we use the configuration of "hbase.master.startup.retainassign" to determine what type of assignment to do on startup,(retainassign or round-robin fashion), however retainassign is not appropriate in the process of ServerShutdownHandler since the old server is already be dead.
> Using bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "Hudson (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13272118#comment-13272118 ]
Hudson commented on HBASE-5914:
-------------------------------
Integrated in HBase-TRUNK-security #197 (See [https://builds.apache.org/job/HBase-TRUNK-security/197/])
HBASE-5914 Bulk assign regions in the process of ServerShutdownHandler (Chunhui) (Revision 1336308)
Result = FAILURE
tedyu :
Files :
* /hbase/trunk/src/main/java/org/apache/hadoop/hbase/master/AssignmentManager.java
* /hbase/trunk/src/main/java/org/apache/hadoop/hbase/master/handler/ServerShutdownHandler.java
* /hbase/trunk/src/test/java/org/apache/hadoop/hbase/master/TestAssignmentManager.java
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch, HBASE-5914v3.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Updated] (HBASE-5914) Bulk assign regions in the process of
ServerShutdownHandler
Posted by "Zhihong Yu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Zhihong Yu updated HBASE-5914:
------------------------------
Fix Version/s: 0.96.0
Summary: Bulk assign regions in the process of ServerShutdownHandler (was: Using bulk assign regions in the process of ServerShutdownHandler)
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Updated] (HBASE-5914) Bulk assign regions in the process of
ServerShutdownHandler
Posted by "Zhihong Ted Yu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Zhihong Ted Yu updated HBASE-5914:
----------------------------------
Resolution: Fixed
Hadoop Flags: Reviewed
Status: Resolved (was: Patch Available)
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch, HBASE-5914v3.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "Zhihong Yu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13271541#comment-13271541 ]
Zhihong Yu commented on HBASE-5914:
-----------------------------------
I was able to get TestDrainingServer pass on MacBook.
Will integrate patch v3 if there is no objection.
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch, HBASE-5914v3.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "chunhui shen (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13267134#comment-13267134 ]
chunhui shen commented on HBASE-5914:
-------------------------------------
bq.how do we know that regions have been assigned ?
We don't need wait until regions assigned, and quickAssign() just ensure the rpc of open region is sent successfully.
For example, in current process of ServerShutdownHandler, it just call AssignmentManager#assign(region, true)(return void) and didn't wait
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "Hudson (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13271657#comment-13271657 ]
Hudson commented on HBASE-5914:
-------------------------------
Integrated in HBase-TRUNK #2858 (See [https://builds.apache.org/job/HBase-TRUNK/2858/])
HBASE-5914 Bulk assign regions in the process of ServerShutdownHandler (Chunhui) (Revision 1336308)
Result = FAILURE
tedyu :
Files :
* /hbase/trunk/src/main/java/org/apache/hadoop/hbase/master/AssignmentManager.java
* /hbase/trunk/src/main/java/org/apache/hadoop/hbase/master/handler/ServerShutdownHandler.java
* /hbase/trunk/src/test/java/org/apache/hadoop/hbase/master/TestAssignmentManager.java
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch, HBASE-5914v3.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "Zhihong Yu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13267158#comment-13267158 ]
Zhihong Yu commented on HBASE-5914:
-----------------------------------
I think round robin is good enough for this issue.
When a concrete non-round robin algorithm is developed, we can consider making that a choice.
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "Enis Soztutar (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13481869#comment-13481869 ]
Enis Soztutar commented on HBASE-5914:
--------------------------------------
[~ram_krish] Ram, do you think that it is safe for backport into 0.94? It seems a good improvement for SSH.
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch, HBASE-5914v3.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "Hadoop QA (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13271214#comment-13271214 ]
Hadoop QA commented on HBASE-5914:
----------------------------------
-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12526119/HBASE-5914v3.patch
against trunk revision .
+1 @author. The patch does not contain any @author tags.
+1 tests included. The patch appears to include 3 new or modified tests.
+1 hadoop23. The patch compiles against the hadoop 0.23.x profile.
+1 javadoc. The javadoc tool did not generate any warning messages.
+1 javac. The applied patch does not increase the total number of javac compiler warnings.
+1 findbugs. The patch does not introduce any new Findbugs (version 1.3.9) warnings.
+1 release audit. The applied patch does not increase the total number of release audit warnings.
-1 core tests. The patch failed these unit tests:
org.apache.hadoop.hbase.TestDrainingServer
Test results: https://builds.apache.org/job/PreCommit-HBASE-Build/1811//testReport/
Findbugs warnings: https://builds.apache.org/job/PreCommit-HBASE-Build/1811//artifact/trunk/patchprocess/newPatchFindbugsWarnings.html
Console output: https://builds.apache.org/job/PreCommit-HBASE-Build/1811//console
This message is automatically generated.
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch, HBASE-5914v3.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Updated] (HBASE-5914) Using bulk assign regions in the
process of ServerShutdownHandler
Posted by "chunhui shen (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
chunhui shen updated HBASE-5914:
--------------------------------
Attachment: HBASE-5914v2.patch
@ted
In the v2 patch, I remove {code}+ if(regions.isEmpty()) return;{code},so quickAssign must assign successfully now.
{code}+ servers.removeAll(failedPlans.keySet());{code}
I think we needn't whether servers is empty, if it is empty, bulkPlan is null, and we will assign region singly.
> Using bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Bulk assign regions in the process
of ServerShutdownHandler
Posted by "ramkrishna.s.vasudevan (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13267164#comment-13267164 ]
ramkrishna.s.vasudevan commented on HBASE-5914:
-----------------------------------------------
Fine Ted. I agree.
> Bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Fix For: 0.96.0
>
> Attachments: HBASE-5914.patch, HBASE-5914v2.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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
[jira] [Commented] (HBASE-5914) Using bulk assign regions in the
process of ServerShutdownHandler
Posted by "Zhihong Yu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-5914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13266578#comment-13266578 ]
Zhihong Yu commented on HBASE-5914:
-----------------------------------
Good idea.
{code}
+ if(regions.isEmpty()) return;
{code}
Please insert space between if and (.
{code}
+ servers.removeAll(failedPlans.keySet());
{code}
I think we should check whether servers is empty before proceeding.
I suggest letting quickAssign() return boolean to indicate whether 'quick' assignment is successful.
> Using bulk assign regions in the process of ServerShutdownHandler
> -----------------------------------------------------------------
>
> Key: HBASE-5914
> URL: https://issues.apache.org/jira/browse/HBASE-5914
> Project: HBase
> Issue Type: Improvement
> Reporter: chunhui shen
> Assignee: chunhui shen
> Attachments: HBASE-5914.patch
>
>
> In the process of ServerShutdownHandler, we currently assign regions singly.
> In the large cluster, one regionserver always carried many regions, this action is quite slow.
> What about using bulk assign regions like cluster start up.
> In current logic, if we failed assigning many regions to one destination server, we will wait unitl timeout,
> however in the process of ServerShutdownHandler, we should retry it to another server.
--
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