[
https://issues.apache.org/jira/browse/HDFS-11594?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15950130#comment-15950130
]
Anu Engineer commented on HDFS-11594:
-------------------------------------
[~cheersyang] Thanks for the comment. This was the same point that [~xyao] also
made. Here some things to consider before we go down that path.
* We have a DB per container. So we might have a million keys per DB. Assuming
we have 10 times more delete operations on some particular open container, even
then I am hopeful that we might not degrade too much in performance.
* The close container ( we have an RPC right now -- mostly for testing and cli
purpose) will most be invoked as a response to a command from SCM. So even if
we take 30 seconds to close a container, we should be ok.
* I am quite open to introducing the compactDB as background thread, we can
have disk scrubber thread do it if needed -- we can find the size of the DB
directory and run compactDB command.
I have filed HDFS-11601 to track this issue. Once more thank you for sharing
your perspective.
> Ozone: close container should call compactDB
> ---------------------------------------------
>
> Key: HDFS-11594
> URL: https://issues.apache.org/jira/browse/HDFS-11594
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: ozone
> Reporter: Anu Engineer
> Assignee: Anu Engineer
> Fix For: HDFS-7240
>
> Attachments: HDFS-11594-HDFS-7240.001.patch
>
>
> Closing a container should call compactDB to compact the LevelDB.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
