[ 
https://issues.apache.org/jira/browse/HDDS-239?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16548289#comment-16548289
 ] 

Xiaoyu Yao commented on HDDS-239:
---------------------------------

Thanks [~msingh] for working on this. The patch looks good to me overall. Here 
are few comments:

 

PipelineManager.java

Line 112: NIT: an extra "pipeline"  in comments

 

Line 127: should we use the lease manager to handle the CLOSING->CLOSED state 
transition when CLOSE event is not received timely? This could be handled as 
TODO in a separate JIRA.

 

 

PipelineSelector.java

Line 72-73: should we move the pipelineLeaseManager/stateMachine to 
PipelineManager class so that the RatisManagerImpl and StandaloneMangerImpl may 
have different state machines? This way, the logic for state transition around 
line 261-266 can be consolidated into PipelineManager#initializePipeline().

 

Also, the createPipeline/closePipeline are not hooked up with state changes 
because we currently rely only getReplicationPipeline() for create pipeline.

 

 

Line 399: I think we need to handle the create timeout by releasing the nodes 
when the DN get closed. Is it possible to include it in this patch? We can 
leave the close pipeline, continer, etc. later.

 

 

ScmConfigKeys.java:

Line 244: please update the unit test TestOzoneConfigurationFields that 
verifies this new configuration keys.

> Add Pipeline StateManager to track and transition pipeline states
> -----------------------------------------------------------------
>
>                 Key: HDDS-239
>                 URL: https://issues.apache.org/jira/browse/HDDS-239
>             Project: Hadoop Distributed Data Store
>          Issue Type: Bug
>          Components: SCM
>    Affects Versions: 0.2.1
>            Reporter: Mukul Kumar Singh
>            Assignee: Mukul Kumar Singh
>            Priority: Major
>             Fix For: 0.2.1
>
>         Attachments: HDDS-239.001.patch, HDDS-239.002.patch, 
> HDDS-239.003.patch
>
>
> With addition of pipeline recovery in Ozone, pipeline failures need to be 
> handled both in Ozone client as well as SCM. This jira adds a pipeline state 
> Manager to manage pipeline state transitions.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org

Reply via email to