hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Daryn Sharp (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-9952) Expose FSNamesystem lock wait time as metrics
Date Wed, 16 Mar 2016 14:21:33 GMT

    [ https://issues.apache.org/jira/browse/HDFS-9952?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15197381#comment-15197381
] 

Daryn Sharp commented on HDFS-9952:
-----------------------------------

Nice.  I've been meaning to add these metrics for quite awhile now.  A few issues:
# I'm not sure if metrics can be null, but I see some places doing the same check.  So if
it can go null, {{NameNode.getNameNodeMetrics()}} must be called once to assign, null checked,
used.
# Write lock wait should only be updated in the "if" for first write lock.  Otherwise reentrant
locking (with near 0 time) will dilute the value.
# Most importantly and adamantly, {{MutableRate#add}} is synchronized in an extremely critical
code path which will destroy concurrent read ops.  Synchronization will not scale with 100
threads doing 20-40k ops/sec, with bursts of 60-100k+ ops/ sec. I'm not a metrics expert but
this must use a non-synchronized metric.

> Expose FSNamesystem lock wait time as metrics
> ---------------------------------------------
>
>                 Key: HDFS-9952
>                 URL: https://issues.apache.org/jira/browse/HDFS-9952
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: namenode
>            Reporter: Vinayakumar B
>            Assignee: Vinayakumar B
>         Attachments: HDFS-9952-01.patch
>
>
> Expose FSNameSystem's readlock() and writeLock() wait time as metrics.



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

Mime
View raw message