[
https://issues.apache.org/jira/browse/MAPREDUCE-2193?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12995776#comment-12995776
]
Chris Douglas commented on MAPREDUCE-2193:
------------------------------------------
bq. I'd like someone else to take a look at adding 'synchronized' to getTasks()
in addition to me - it's just scary enough that I think an extra reviewer is
prudent.
I tried to find some history: http://s.apache.org/T99 (from MAPREDUCE-1316).
The synchronization was intentionally left out, but only to be consistent with
the existing code.
The synchronization around JIP is difficult to track. Its omission here appears
to admit the possibility that a query of an initializing job could iterate over
a partially constructed set of TIPs and cause NPEs (e.g. via the JSP), but I
couldn't find a place where this is a serious problem. After the job is
initialized, none of the references to member fields become stale. On the other
hand, I looked through all the callers of getTasks(), and most are holding the
JT lock already (mostly schedulers, during heartbeat processing) so I doubt
that synchronizing on the JIP here would cause problems.
One could fix getTasks() to be threadsafe by changing create{Map,Reduce}Tasks
and initSetupCleanupTasks to assign the initialized array of tasks to the
member field. The current behavior constructs it in-place, which (AFAICT) means
that a caller of getTasks() could read partially-constructed TIPs. It's
probably better to make getTasks() synchronized, to avoid having getTasks()
return partially constructed, inconsistent JIP state.
> 13 Findbugs warnings on trunk and branch-0.22
> ---------------------------------------------
>
> Key: MAPREDUCE-2193
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-2193
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Affects Versions: 0.22.0, 0.23.0
> Reporter: Nigel Daley
> Assignee: Harsh J Chouraria
> Priority: Blocker
> Fix For: 0.22.0, 0.23.0
>
> Attachments: findbugsWarnings.html, hadoop-findbugs-report.html,
> hadoop-findbugs-report.html, mapreduce.trunk.findbugs.r1.diff,
> mapreduce.trunk.findbugs.r2.diff, mapreduce.trunk.findbugs.r3.diff,
> mapreduce.trunk.findbugs.r4.diff
>
>
> There are 13 findbugs warnings on trunk. See attached html file. These must
> be fixed or filtered out to get back to 0 warnings. The OK_FINDBUGS_WARNINGS
> property in src/test/test-patch.properties should also be set to 0 in the
> patch that fixes this issue.
--
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
