You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "stack (JIRA)" <ji...@apache.org> on 2010/06/29 03:59:49 UTC
[jira] Created: (HBASE-2802) Stop daughter regions both being
assigned same server
Stop daughter regions both being assigned same server
-----------------------------------------------------
Key: HBASE-2802
URL: https://issues.apache.org/jira/browse/HBASE-2802
Project: HBase
Issue Type: Improvement
Reporter: stack
Fix For: 0.20.6, 0.21.0
Attachments: moz.txt
On split, usually the two daughter regions end up on assignment list at same time. Dependent on cluster state and what balancer says, when a server checks in, we'll assign at least two regions in the one go so daughters then end up on same server. This patch makes it so we don't assign contiguous regions in the one assignment.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Resolved: (HBASE-2802) Stop daughter regions both being
assigned same server
Posted by "stack (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-2802?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
stack resolved HBASE-2802.
--------------------------
Hadoop Flags: [Reviewed]
Fix Version/s: (was: 0.90.0)
Resolution: Fixed
Committed to 0.20 branch.
> Stop daughter regions both being assigned same server
> -----------------------------------------------------
>
> Key: HBASE-2802
> URL: https://issues.apache.org/jira/browse/HBASE-2802
> Project: HBase
> Issue Type: Improvement
> Reporter: stack
> Fix For: 0.20.6
>
> Attachments: moz.txt
>
>
> On split, usually the two daughter regions end up on assignment list at same time. Dependent on cluster state and what balancer says, when a server checks in, we'll assign at least two regions in the one go so daughters then end up on same server. This patch makes it so we don't assign contiguous regions in the one assignment.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HBASE-2802) Stop daughter regions both being
assigned same server
Posted by "Jean-Daniel Cryans (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-2802?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12884846#action_12884846 ]
Jean-Daniel Cryans commented on HBASE-2802:
-------------------------------------------
IMO this issue is a workaround for the master's inability to balance on requests load (instead of regions). With better load balancing, we should do a better job at redistributing regions on the cluster so this patch won't be needed.
> Stop daughter regions both being assigned same server
> -----------------------------------------------------
>
> Key: HBASE-2802
> URL: https://issues.apache.org/jira/browse/HBASE-2802
> Project: HBase
> Issue Type: Improvement
> Reporter: stack
> Fix For: 0.20.6, 0.21.0
>
> Attachments: moz.txt
>
>
> On split, usually the two daughter regions end up on assignment list at same time. Dependent on cluster state and what balancer says, when a server checks in, we'll assign at least two regions in the one go so daughters then end up on same server. This patch makes it so we don't assign contiguous regions in the one assignment.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HBASE-2802) Stop daughter regions both being
assigned same server
Posted by "stack (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-2802?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
stack updated HBASE-2802:
-------------------------
Attachment: moz.txt
> Stop daughter regions both being assigned same server
> -----------------------------------------------------
>
> Key: HBASE-2802
> URL: https://issues.apache.org/jira/browse/HBASE-2802
> Project: HBase
> Issue Type: Improvement
> Reporter: stack
> Fix For: 0.20.6, 0.21.0
>
> Attachments: moz.txt
>
>
> On split, usually the two daughter regions end up on assignment list at same time. Dependent on cluster state and what balancer says, when a server checks in, we'll assign at least two regions in the one go so daughters then end up on same server. This patch makes it so we don't assign contiguous regions in the one assignment.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HBASE-2802) Stop daughter regions both being
assigned same server
Posted by "stack (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-2802?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12890173#action_12890173 ]
stack commented on HBASE-2802:
------------------------------
In cluster testing I'll see messages like the below:
{code}
2010-07-19 14:14:48,065 INFO org.apache.hadoop.hbase.master.ServerManager: Processing MSG_REPORT_SPLIT_INCLUDES_DAUGHTERS: TestTable,0073837150,1279573828629: Daughters; TestTable,0073837150,1279574087136, TestTable,0073971166,1279574087136 from sv2borg188,60020,1279565949537; 1 of 1
2010-07-19 14:14:48,068 DEBUG org.apache.hadoop.hbase.master.RegionManager: Assigning for address: 10.20.20.188:60020, startcode: 1279565949537, load: (requests=0, regions=114, usedHeap=2183, maxHeap=3045): total nregions to assign=2, regions to give other servers than this=0, isMetaAssign=false
2010-07-19 14:14:48,068 DEBUG org.apache.hadoop.hbase.master.RegionManager: Assigning address: 10.20.20.188:60020, startcode: 1279565949537, load: (requests=0, regions=114, usedHeap=2183, maxHeap=3045) 2 regions
2010-07-19 14:14:48,068 INFO org.apache.hadoop.hbase.master.RegionManager: Assigning region TestTable,0073837150,1279574087136 to sv2borg188,60020,1279565949537
2010-07-19 14:14:48,068 INFO org.apache.hadoop.hbase.master.RegionManager: Passing on TestTable,0073971166,1279574087136 because contiguous with TestTable,0073837150,1279574087136 (will assign next server that checks in)
2010-07-19 14:14:48,184 DEBUG org.apache.hadoop.hbase.master.RegionManager: Assigning for address: 10.20.20.187:60020, startcode: 1279565949563, load: (requests=0, regions=114, usedHeap=2166, maxHeap=3045): total nregions to assign=1, regions to give other servers than this=0, isMetaAssign=false
2010-07-19 14:14:48,184 DEBUG org.apache.hadoop.hbase.master.RegionManager: Assigning address: 10.20.20.187:60020, startcode: 1279565949563, load: (requests=0, regions=114, usedHeap=2166, maxHeap=3045) 1 regions
2010-07-19 14:14:48,184 INFO org.apache.hadoop.hbase.master.RegionManager: Assigning region TestTable,0073971166,1279574087136 to sv2borg187,60020,1279565949563
{code}
It seems to be working.
> Stop daughter regions both being assigned same server
> -----------------------------------------------------
>
> Key: HBASE-2802
> URL: https://issues.apache.org/jira/browse/HBASE-2802
> Project: HBase
> Issue Type: Improvement
> Reporter: stack
> Fix For: 0.20.6
>
> Attachments: moz.txt
>
>
> On split, usually the two daughter regions end up on assignment list at same time. Dependent on cluster state and what balancer says, when a server checks in, we'll assign at least two regions in the one go so daughters then end up on same server. This patch makes it so we don't assign contiguous regions in the one assignment.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HBASE-2802) Stop daughter regions both being
assigned same server
Posted by "Jonathan Gray (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-2802?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12884844#action_12884844 ]
Jonathan Gray commented on HBASE-2802:
--------------------------------------
So then we do not want to go through with other split changes that have regions re-open on server hosting parent?
> Stop daughter regions both being assigned same server
> -----------------------------------------------------
>
> Key: HBASE-2802
> URL: https://issues.apache.org/jira/browse/HBASE-2802
> Project: HBase
> Issue Type: Improvement
> Reporter: stack
> Fix For: 0.20.6, 0.21.0
>
> Attachments: moz.txt
>
>
> On split, usually the two daughter regions end up on assignment list at same time. Dependent on cluster state and what balancer says, when a server checks in, we'll assign at least two regions in the one go so daughters then end up on same server. This patch makes it so we don't assign contiguous regions in the one assignment.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HBASE-2802) Stop daughter regions both being
assigned same server
Posted by "stack (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-2802?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12884864#action_12884864 ]
stack commented on HBASE-2802:
------------------------------
To be clear this issue is inthe main targeted at 0. 20 branch to help out our Mozilla brethren
> Stop daughter regions both being assigned same server
> -----------------------------------------------------
>
> Key: HBASE-2802
> URL: https://issues.apache.org/jira/browse/HBASE-2802
> Project: HBase
> Issue Type: Improvement
> Reporter: stack
> Fix For: 0.20.6, 0.21.0
>
> Attachments: moz.txt
>
>
> On split, usually the two daughter regions end up on assignment list at same time. Dependent on cluster state and what balancer says, when a server checks in, we'll assign at least two regions in the one go so daughters then end up on same server. This patch makes it so we don't assign contiguous regions in the one assignment.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.