[jira] [Commented] (HDFS-7964) Add support for async edit logging

["Daryn Sharp (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HDFS-7964?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14701823#comment-14701823 ] 

Daryn Sharp commented on HDFS-7964:
-----------------------------------

[~hitliuyi], I'm sure I'll need to rebase.

The batched metric is the count of edits being implicitly sync'ed by another edit, not the sum of all edits in the sync.  That was the pre-existing semantic I didn't want to alter.

Technically AsyncEdit#logSyncNotify doesn't need sync, but findbugs gets grumbly about access to {{done}}.

Re: conf.  At the time, we didn't have much run time on it so I considered it experimental.  All 2.6 clusters have been running with the feature enabled for 6+ months so I'll add to the default conf.

> Add support for async edit logging
> ----------------------------------
>
>                 Key: HDFS-7964
>                 URL: https://issues.apache.org/jira/browse/HDFS-7964
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: namenode
>    Affects Versions: 2.0.2-alpha
>            Reporter: Daryn Sharp
>            Assignee: Daryn Sharp
>         Attachments: HDFS-7964.patch
>
>
> Edit logging is a major source of contention within the NN.  LogEdit is called within the namespace write log, while logSync is called outside of the lock to allow greater concurrency.  The handler thread remains busy until logSync returns to provide the client with a durability guarantee for the response.
> Write heavy RPC load and/or slow IO causes handlers to stall in logSync.  Although the write lock is not held, readers are limited/starved and the call queue fills.  Combining an edit log thread with postponed RPC responses from HADOOP-10300 will provide the same durability guarantee but immediately free up the handlers.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)