Wangda Tan commented on YARN-3388:

Thanks for updating, I took a look at your patch, approach LGTM, but I think 
node label should be considered in the same JIRA, when doing allocation under a 
labeled node, user-limit checking in the patch is incorrect. Actually 
user-limit for exclusive node label is already supported in latest trunk, and 
user-limit for non-exclusive node label is contains in YARN-3361, I think after 
YARN-3361, user-limit for node label will be in a good shape.
Would you mind to take a look at computeUserLimit method of the patch attached 
in YARN-3361?

To support computing consumed-per-partition, User.updateUsageRatio need receive 
partition as parameter.

Some other comments:
- consumedRatio -> totalDominateConsumed or some other name, it's better to 
make sum of "dominate" consumed in the name
- consumed -> totalDominatedConsumedByPartition.
- It's better to add a test case to make sure allocation locked descripted in 
 will not happened.

> Allocation in LeafQueue could get stuck because DRF calculator isn't well 
> supported when computing user-limit
> -------------------------------------------------------------------------------------------------------------
>                 Key: YARN-3388
>                 URL: https://issues.apache.org/jira/browse/YARN-3388
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: capacityscheduler
>    Affects Versions: 2.6.0
>            Reporter: Nathan Roberts
>            Assignee: Nathan Roberts
>         Attachments: YARN-3388-v0.patch, YARN-3388-v1.patch
> When there are multiple active users in a queue, it should be possible for 
> those users to make use of capacity up-to max_capacity (or close). The 
> resources should be fairly distributed among the active users in the queue. 
> This works pretty well when there is a single resource being scheduled.   
> However, when there are multiple resources the situation gets more complex 
> and the current algorithm tends to get stuck at Capacity. 
> Example illustrated in subsequent comment.

This message was sent by Atlassian JIRA

Reply via email to