[ 
https://issues.apache.org/jira/browse/YARN-2608?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14148354#comment-14148354
 ] 

Karthik Kambatla commented on YARN-2608:
----------------------------------------

+1, pending Jenkins. 

> FairScheduler may hung due to two potential deadlocks
> -----------------------------------------------------
>
>                 Key: YARN-2608
>                 URL: https://issues.apache.org/jira/browse/YARN-2608
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Wei Yan
>            Assignee: Wei Yan
>         Attachments: YARN-2608-1.patch, YARN-2608-2.patch
>
>
> Two potential deadlocks exist inside the FairScheduler.
> 1. AllocationFileLoaderService would reload the queue configuration, which 
> calls FairScheduler.AllocationReloadListener.onReload() function. And require 
> *FairScheduler's lock*; 
> {code}
>   public void onReload(AllocationConfiguration queueInfo) {
>       synchronized (FairScheduler.this) {
>           ....
>       }
>   }
> {code}
> after that, it would require the *QueueManager's queues lock*.
> {code}
>   private FSQueue getQueue(String name, boolean create, FSQueueType 
> queueType) {
>       name = ensureRootPrefix(name);
>       synchronized (queues) {
>           ....
>       }
>   }
> {code}
> Another thread FairScheduler.assignToQueue may also need to create a new 
> queue when a new job submitted. This thread would hold the *QueueManager's 
> queues lock* firstly, and then would like to hold the *FairScheduler's lock* 
> as it needs to call FairScheduler.getClock() function when creating a new 
> FSLeafQueue. Deadlock may happen here.
> 2. The AllocationFileLoaderService holds  *AllocationFileLoaderService's 
> lock* first, and then waits for *FairScheduler's lock*. Another thread (like 
> AdminService.refreshQueues) may call FairScheduler's reinitialize function, 
> which holds *FairScheduler's lock* first, and then waits for 
> *AllocationFileLoaderService's lock*. Deadlock may happen here.



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

Reply via email to