[
https://issues.apache.org/jira/browse/HDFS-14657?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16887596#comment-16887596
]
Chen Zhang commented on HDFS-14657:
-----------------------------------
Thanks [~shv] for your comments, I did searched for a few jiras related with
the block-reports, but unfortunately I missed the HDFS-11313 you mentioned above
I just go though all the discussion under the jira HDFS-11313, looks the main
concern is the race conditions between SBR and IBR, this problem is considered
in my solution:
# Add a report lock to DatanodeDescriptor
# Before processing the FBR and IBR, BlockManager should get the report lock
for that node first
# IBR must wait until FBR process complete, even the writelock may release and
re-acquire many times during processing FBR
This solution is quite simple and already deployed to our largest production
cluster for more than 1 year, it's very stable.
I think SBR is a good idea and it looks more elegant, is there anyone still
pushing that forward?
+[~daryn], you have given a lot of valuable advice for HDFS-11313, what's your
opinion?
> Refine NameSystem lock usage during processing FBR
> --------------------------------------------------
>
> Key: HDFS-14657
> URL: https://issues.apache.org/jira/browse/HDFS-14657
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: Chen Zhang
> Assignee: Chen Zhang
> Priority: Major
> Attachments: HDFS-14657-001.patch
>
>
> The disk with 12TB capacity is very normal today, which means the FBR size is
> much larger than before, Namenode holds the NameSystemLock during processing
> block report for each storage, which might take quite a long time.
> On our production environment, processing large FBR usually cause a longer
> RPC queue time, which impacts client latency, so we did some simple work on
> refining the lock usage, which improved the p99 latency significantly.
> In our solution, BlockManager release the NameSystem write lock and request
> it again for every 5000 blocks(by default) during processing FBR, with the
> fair lock, all the RPC request can be processed before BlockManager
> re-acquire the write lock.
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
