hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "zhihai xu (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-3204) Fix new findbug warnings in hadoop-yarn-server-resourcemanager(resourcemanager.scheduler.fair)
Date Tue, 17 Mar 2015 19:11:41 GMT

    [ https://issues.apache.org/jira/browse/YARN-3204?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14365821#comment-14365821
] 

zhihai xu commented on YARN-3204:
---------------------------------

Some comments:
1. about the comment for Inconsistent sync warning fsOpDurations
{code}
Inconsistent sync warning - callDurationMetrics is only initialized once and never changed
{code}
It looks like not accurate. Each method from fsOpDurations is only called in one thread, all
these methods access different fields and are independent.

2.Can we define reloadListener as volatile?
since reloadListener is accessed by two threads, it will be safer to use volatile.

3.Can we move the check to the beginning of the functions of reserveResource?
It will be better to check error earlier than later to avoid unnecessary operations.
{code}
if (!(application instanceof FSAppAttempt)) {
{code}
Can we use YarnRuntimeException instead of IllegalArgumentException?
This looks like an unexpected runtime exception.

4.adding lock for getAllocationConfiguration is dangerous.
A lot of codes(Queue, FairReservationSystem... ) are calling getAllocationConfiguration, which
can introduce potential deadlock situation and performance issue.
For example, QueueManager#getQueue lock queues then call QueueManager#createQueue then call
scheduler.getAllocationConfiguration.
This will have two layer locks if we add lock in getAllocationConfiguration.
Can we define allocConf as volatile?  since allocConf will only be updated by AllocationReloadListener.onReload
which is called from AllocationFileLoaderService#reloadThread after initialization. 


> Fix new findbug warnings in hadoop-yarn-server-resourcemanager(resourcemanager.scheduler.fair)
> ----------------------------------------------------------------------------------------------
>
>                 Key: YARN-3204
>                 URL: https://issues.apache.org/jira/browse/YARN-3204
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Brahma Reddy Battula
>            Assignee: Brahma Reddy Battula
>            Priority: Blocker
>         Attachments: YARN-3204-001.patch, YARN-3204-002.patch, YARN-3204-003.patch
>
>
> Please check following findbug report..
> https://builds.apache.org/job/PreCommit-YARN-Build/6644//artifact/patchprocess/newPatchFindbugsWarningshadoop-yarn-server-resourcemanager.html



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

Mime
View raw message