[ 
https://issues.apache.org/jira/browse/HDFS-14593?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16882384#comment-16882384
 ] 

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

Thanks, [~elgoiri].

bq. My only concern now is the fact that we are going to the backend (say 
ZooKeeper) twice, right?

Yes, you are right. It's definitely a problem.

Uploaded 007.patch. It refers the backend once, and it removes expired records 
from both the cache and the backend in {{overrideExpiredRecord()}}.

To modify records in the cache, it includes two additional changes.
* use ArrayList instead of Collections.singletonList in 
{{CachedRecordStore#overrideExpiredRecord(R record)}}
* not use Collections.unmodifiableList in {{QueryResult#getRecords()}}

I'm not sure whether the changes of 007.patch can be acceptable. If not, it may 
be better just not to change the order in {{CachedRecordStore#loadCache()}}. We 
may be able to compromise the behavior that I mentioned before.

bq. If not changing the order, expired records keep existing in the 
cache(result at CachedRecordStore#118) after deleting them from State Store by 
overrideExpiredRecords(). Then, they are deleted from the cache in the next 
loadCache.


> RBF: Implement deletion feature for expired records in State Store
> ------------------------------------------------------------------
>
>                 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, HDFS-14593.004.patch, HDFS-14593.005.patch, 
> HDFS-14593.006.patch, HDFS-14593.007.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