You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Gregory Chanan (JIRA)" <ji...@apache.org> on 2012/11/16 00:03:12 UTC
[jira] [Created] (HBASE-7170) [0.94 branch] Allow
HConnectionImplementation to reconnect to master multiple times
Gregory Chanan created HBASE-7170:
-------------------------------------
Summary: [0.94 branch] Allow HConnectionImplementation to reconnect to master multiple times
Key: HBASE-7170
URL: https://issues.apache.org/jira/browse/HBASE-7170
Project: HBase
Issue Type: Task
Reporter: Gregory Chanan
Assignee: Gregory Chanan
Priority: Minor
Fix For: 0.94.4
HBASE-5058 retained the old behavior of managed HConnections not retrying their connection to master b/c of the "masterChecked" variable.
This behavior isn't great though, because you can get in the following situation:
1) Connect to master using HBaseAdmin
2) Master goes down
3) New master takes over
4) The HBaseAdmin object is now useless
The client needs to create a new HBaseAdmin in this case.
--
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-7170) [0.94 branch] Allow
HConnectionImplementation to reconnect to master multiple times
Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-7170?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498554#comment-13498554 ]
Lars Hofhansl commented on HBASE-7170:
--------------------------------------
+1 on removing that masterChecked nonsense.
Does anybody know why it was there in the first place? (I have a bit of a nagging feeling that it was put in for a reason)
> [0.94 branch] Allow HConnectionImplementation to reconnect to master multiple times
> -----------------------------------------------------------------------------------
>
> Key: HBASE-7170
> URL: https://issues.apache.org/jira/browse/HBASE-7170
> Project: HBase
> Issue Type: Task
> Reporter: Gregory Chanan
> Assignee: Gregory Chanan
> Priority: Minor
> Fix For: 0.94.4
>
> Attachments: HBASE-7170.patch
>
>
> HBASE-5058 retained the old behavior of managed HConnections not retrying their connection to master b/c of the "masterChecked" variable.
> This behavior isn't great though, because you can get in the following situation:
> 1) Connect to master using HBaseAdmin
> 2) Master goes down
> 3) New master takes over
> 4) The HBaseAdmin object is now useless
> The client needs to create a new HBaseAdmin in this case.
--
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-7170) [0.94 branch] Allow
HConnectionImplementation to reconnect to master multiple times
Posted by "stack (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-7170?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498474#comment-13498474 ]
stack commented on HBASE-7170:
------------------------------
TestFromClientSide intentionally has lots of tests in it to cut down on our spinning up cluster instances.
Patch looks good to me. You need to release note changed behavior?
> [0.94 branch] Allow HConnectionImplementation to reconnect to master multiple times
> -----------------------------------------------------------------------------------
>
> Key: HBASE-7170
> URL: https://issues.apache.org/jira/browse/HBASE-7170
> Project: HBase
> Issue Type: Task
> Reporter: Gregory Chanan
> Assignee: Gregory Chanan
> Priority: Minor
> Fix For: 0.94.4
>
> Attachments: HBASE-7170.patch
>
>
> HBASE-5058 retained the old behavior of managed HConnections not retrying their connection to master b/c of the "masterChecked" variable.
> This behavior isn't great though, because you can get in the following situation:
> 1) Connect to master using HBaseAdmin
> 2) Master goes down
> 3) New master takes over
> 4) The HBaseAdmin object is now useless
> The client needs to create a new HBaseAdmin in this case.
--
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] [Updated] (HBASE-7170) [0.94 branch] Allow
HConnectionImplementation to reconnect to master multiple times
Posted by "Gregory Chanan (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-7170?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Gregory Chanan updated HBASE-7170:
----------------------------------
Attachment: HBASE-7170.patch
Here's a patch that illustrates the problem and fixes it.
This backports HBASE-6606 (the testUnmanagedHConnectionReconnect test in TestFromClientSide) to 0.94. The test causes the master to die and a new master to come back up in TestFromClientSide.
But due to HBASE-5051, all the tests in TestFromClientSide use the same HBaseAdmin, so all the tests that are executed after testUnmanagedHConnectionReconnect fail.
> [0.94 branch] Allow HConnectionImplementation to reconnect to master multiple times
> -----------------------------------------------------------------------------------
>
> Key: HBASE-7170
> URL: https://issues.apache.org/jira/browse/HBASE-7170
> Project: HBase
> Issue Type: Task
> Reporter: Gregory Chanan
> Assignee: Gregory Chanan
> Priority: Minor
> Fix For: 0.94.4
>
> Attachments: HBASE-7170.patch
>
>
> HBASE-5058 retained the old behavior of managed HConnections not retrying their connection to master b/c of the "masterChecked" variable.
> This behavior isn't great though, because you can get in the following situation:
> 1) Connect to master using HBaseAdmin
> 2) Master goes down
> 3) New master takes over
> 4) The HBaseAdmin object is now useless
> The client needs to create a new HBaseAdmin in this case.
--
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-7170) [0.94 branch] Allow
HConnectionImplementation to reconnect to master multiple times
Posted by "Lars Hofhansl (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-7170?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13499223#comment-13499223 ]
Lars Hofhansl commented on HBASE-7170:
--------------------------------------
Awesome, thanks Gregory. Let's just remove it.
> [0.94 branch] Allow HConnectionImplementation to reconnect to master multiple times
> -----------------------------------------------------------------------------------
>
> Key: HBASE-7170
> URL: https://issues.apache.org/jira/browse/HBASE-7170
> Project: HBase
> Issue Type: Task
> Reporter: Gregory Chanan
> Assignee: Gregory Chanan
> Priority: Minor
> Fix For: 0.94.4
>
> Attachments: HBASE-7170.patch
>
>
> HBASE-5058 retained the old behavior of managed HConnections not retrying their connection to master b/c of the "masterChecked" variable.
> This behavior isn't great though, because you can get in the following situation:
> 1) Connect to master using HBaseAdmin
> 2) Master goes down
> 3) New master takes over
> 4) The HBaseAdmin object is now useless
> The client needs to create a new HBaseAdmin in this case.
--
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-7170) [0.94 branch] Allow
HConnectionImplementation to reconnect to master multiple times
Posted by "Gregory Chanan (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-7170?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13499219#comment-13499219 ]
Gregory Chanan commented on HBASE-7170:
---------------------------------------
I traced its history, it was in the initial commit of HConnectionManager:
https://github.com/apache/hbase/commit/07657c40e9528b32bb17f2af72f3b3c34eeada7f
so no additional information on why it was there in the first place.
> [0.94 branch] Allow HConnectionImplementation to reconnect to master multiple times
> -----------------------------------------------------------------------------------
>
> Key: HBASE-7170
> URL: https://issues.apache.org/jira/browse/HBASE-7170
> Project: HBase
> Issue Type: Task
> Reporter: Gregory Chanan
> Assignee: Gregory Chanan
> Priority: Minor
> Fix For: 0.94.4
>
> Attachments: HBASE-7170.patch
>
>
> HBASE-5058 retained the old behavior of managed HConnections not retrying their connection to master b/c of the "masterChecked" variable.
> This behavior isn't great though, because you can get in the following situation:
> 1) Connect to master using HBaseAdmin
> 2) Master goes down
> 3) New master takes over
> 4) The HBaseAdmin object is now useless
> The client needs to create a new HBaseAdmin in this case.
--
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-7170) [0.94 branch] Allow
HConnectionImplementation to reconnect to master multiple times
Posted by "Hudson (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-7170?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13499307#comment-13499307 ]
Hudson commented on HBASE-7170:
-------------------------------
Integrated in HBase-0.94 #598 (See [https://builds.apache.org/job/HBase-0.94/598/])
HBASE-7170 Allow HConnectionImplementation to reconnect to master multiple times (Revision 1410622)
Result = SUCCESS
gchanan :
Files :
* /hbase/branches/0.94/src/main/java/org/apache/hadoop/hbase/client/HConnectionManager.java
* /hbase/branches/0.94/src/test/java/org/apache/hadoop/hbase/client/TestFromClientSide.java
> [0.94 branch] Allow HConnectionImplementation to reconnect to master multiple times
> -----------------------------------------------------------------------------------
>
> Key: HBASE-7170
> URL: https://issues.apache.org/jira/browse/HBASE-7170
> Project: HBase
> Issue Type: Task
> Reporter: Gregory Chanan
> Assignee: Gregory Chanan
> Priority: Minor
> Fix For: 0.94.4
>
> Attachments: HBASE-7170.patch
>
>
> HBASE-5058 retained the old behavior of managed HConnections not retrying their connection to master b/c of the "masterChecked" variable.
> This behavior isn't great though, because you can get in the following situation:
> 1) Connect to master using HBaseAdmin
> 2) Master goes down
> 3) New master takes over
> 4) The HBaseAdmin object is now useless
> The client needs to create a new HBaseAdmin in this case.
--
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-7170) [0.94 branch] Allow
HConnectionImplementation to reconnect to master multiple times
Posted by "stack (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-7170?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13499226#comment-13499226 ]
stack commented on HBASE-7170:
------------------------------
5 years ago!
> [0.94 branch] Allow HConnectionImplementation to reconnect to master multiple times
> -----------------------------------------------------------------------------------
>
> Key: HBASE-7170
> URL: https://issues.apache.org/jira/browse/HBASE-7170
> Project: HBase
> Issue Type: Task
> Reporter: Gregory Chanan
> Assignee: Gregory Chanan
> Priority: Minor
> Fix For: 0.94.4
>
> Attachments: HBASE-7170.patch
>
>
> HBASE-5058 retained the old behavior of managed HConnections not retrying their connection to master b/c of the "masterChecked" variable.
> This behavior isn't great though, because you can get in the following situation:
> 1) Connect to master using HBaseAdmin
> 2) Master goes down
> 3) New master takes over
> 4) The HBaseAdmin object is now useless
> The client needs to create a new HBaseAdmin in this case.
--
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-7170) [0.94 branch] Allow
HConnectionImplementation to reconnect to master multiple times
Posted by "Gregory Chanan (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-7170?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498427#comment-13498427 ]
Gregory Chanan commented on HBASE-7170:
---------------------------------------
Two notes:
- the last sentence should be "all the tests that are executed after testUnmanagedHConnectionReconnect fail [without the changes to HConnectionImplementation]"
- Trunk doesn't have this issue.
> [0.94 branch] Allow HConnectionImplementation to reconnect to master multiple times
> -----------------------------------------------------------------------------------
>
> Key: HBASE-7170
> URL: https://issues.apache.org/jira/browse/HBASE-7170
> Project: HBase
> Issue Type: Task
> Reporter: Gregory Chanan
> Assignee: Gregory Chanan
> Priority: Minor
> Fix For: 0.94.4
>
> Attachments: HBASE-7170.patch
>
>
> HBASE-5058 retained the old behavior of managed HConnections not retrying their connection to master b/c of the "masterChecked" variable.
> This behavior isn't great though, because you can get in the following situation:
> 1) Connect to master using HBaseAdmin
> 2) Master goes down
> 3) New master takes over
> 4) The HBaseAdmin object is now useless
> The client needs to create a new HBaseAdmin in this case.
--
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-7170) [0.94 branch] Allow
HConnectionImplementation to reconnect to master multiple times
Posted by "Ted Yu (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-7170?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498439#comment-13498439 ]
Ted Yu commented on HBASE-7170:
-------------------------------
+1 on patch.
I think Enis found the same problem in HBASE-7009.
{code}
Running org.apache.hadoop.hbase.client.TestFromClientSide
2012-11-15 15:18:44.087 java[71352:1903] Unable to load realm mapping info from SCDynamicStore
Tests run: 59, Failures: 0, Errors: 0, Skipped: 3, Time elapsed: 203.407 sec
{code}
Can we separate some tests out of TestFromClientSide ? In trunk, there is TestFromClientSide3.java
> [0.94 branch] Allow HConnectionImplementation to reconnect to master multiple times
> -----------------------------------------------------------------------------------
>
> Key: HBASE-7170
> URL: https://issues.apache.org/jira/browse/HBASE-7170
> Project: HBase
> Issue Type: Task
> Reporter: Gregory Chanan
> Assignee: Gregory Chanan
> Priority: Minor
> Fix For: 0.94.4
>
> Attachments: HBASE-7170.patch
>
>
> HBASE-5058 retained the old behavior of managed HConnections not retrying their connection to master b/c of the "masterChecked" variable.
> This behavior isn't great though, because you can get in the following situation:
> 1) Connect to master using HBaseAdmin
> 2) Master goes down
> 3) New master takes over
> 4) The HBaseAdmin object is now useless
> The client needs to create a new HBaseAdmin in this case.
--
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] [Resolved] (HBASE-7170) [0.94 branch] Allow
HConnectionImplementation to reconnect to master multiple times
Posted by "Gregory Chanan (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-7170?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Gregory Chanan resolved HBASE-7170.
-----------------------------------
Resolution: Fixed
Thanks for the reviews, committed to 0.94.
> [0.94 branch] Allow HConnectionImplementation to reconnect to master multiple times
> -----------------------------------------------------------------------------------
>
> Key: HBASE-7170
> URL: https://issues.apache.org/jira/browse/HBASE-7170
> Project: HBase
> Issue Type: Task
> Reporter: Gregory Chanan
> Assignee: Gregory Chanan
> Priority: Minor
> Fix For: 0.94.4
>
> Attachments: HBASE-7170.patch
>
>
> HBASE-5058 retained the old behavior of managed HConnections not retrying their connection to master b/c of the "masterChecked" variable.
> This behavior isn't great though, because you can get in the following situation:
> 1) Connect to master using HBaseAdmin
> 2) Master goes down
> 3) New master takes over
> 4) The HBaseAdmin object is now useless
> The client needs to create a new HBaseAdmin in this case.
--
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-7170) [0.94 branch] Allow
HConnectionImplementation to reconnect to master multiple times
Posted by "Enis Soztutar (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-7170?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13499249#comment-13499249 ]
Enis Soztutar commented on HBASE-7170:
--------------------------------------
Thanks Gregory, as Ted pointed out we did run into this in HBASE-7009, and I was going to file the jira, but you beat me to it.
> [0.94 branch] Allow HConnectionImplementation to reconnect to master multiple times
> -----------------------------------------------------------------------------------
>
> Key: HBASE-7170
> URL: https://issues.apache.org/jira/browse/HBASE-7170
> Project: HBase
> Issue Type: Task
> Reporter: Gregory Chanan
> Assignee: Gregory Chanan
> Priority: Minor
> Fix For: 0.94.4
>
> Attachments: HBASE-7170.patch
>
>
> HBASE-5058 retained the old behavior of managed HConnections not retrying their connection to master b/c of the "masterChecked" variable.
> This behavior isn't great though, because you can get in the following situation:
> 1) Connect to master using HBaseAdmin
> 2) Master goes down
> 3) New master takes over
> 4) The HBaseAdmin object is now useless
> The client needs to create a new HBaseAdmin in this case.
--
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