[ 
https://issues.apache.org/jira/browse/HDFS-14593?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16878334#comment-16878334
 ] 

Takanobu Asanuma commented on HDFS-14593:
-----------------------------------------

About 003.patch:
 # To keep consistent between NN and Router, the patch allows to delete expired 
namenode membership too.
 # The default deletion time is negative(disabled).
 # In {{CachedRecordStore}}, {{newRecords}} should be set after 
{{overrideExpiredRecords()}}. If not, expired records remain in the cache after 
deleting them.
 # In {{RouterStatePBImpl}}, the patch skips updating the modification time 
when the record status is EXPIRED. I think it should do that because:
 ** Currently expired records are periodically updated as EXPIRED and the 
modification time is also periodically updated. In that case, it is difficult 
to understand how long the expired record exists and whether it should be 
deleted or not.
 ** In Routers Information Web UI, even if a router is dead(EXPIRED), its 
heartbeat is periodically updated because heartbeat is based on the 
modification time.

> RBF: RouterAdmin should be able to remove expired routers from Routers 
> Information
> ----------------------------------------------------------------------------------
>
>                 Key: HDFS-14593
>                 URL: https://issues.apache.org/jira/browse/HDFS-14593
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: rbf
>            Reporter: Takanobu Asanuma
>            Assignee: Takanobu Asanuma
>            Priority: Major
>         Attachments: HDFS-14593.001.patch, HDFS-14593.002.patch, 
> HDFS-14593.003.patch
>
>
> Currently, any router seems to exist in the Router Information eternally.



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

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to