[jira] [Commented] (YARN-6052) Yarn RM UI % of Queue at application level is wrong
[ https://issues.apache.org/jira/browse/YARN-6052?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15803874#comment-15803874 ] Sunil G commented on YARN-6052: --- Thanks [~Prabhu Joseph] for pointing out. > Yarn RM UI % of Queue at application level is wrong > --- > > Key: YARN-6052 > URL: https://issues.apache.org/jira/browse/YARN-6052 > Project: Hadoop YARN > Issue Type: Bug >Affects Versions: 2.7.3 >Reporter: Prabhu Joseph >Assignee: Prabhu Joseph >Priority: Minor > Attachments: RM_UI.png > > > Test Case: > yarn.scheduler.capacity.root.capacity=100 > yarn.scheduler.capacity.root.queues=default,dummy > yarn.scheduler.capacity.root.default.capacity=20 > yarn.scheduler.capacity.root.dummy.capacity=80 > yarn.scheduler.capacity.root.dummy.child.capacity=50 > yarn.scheduler.capacity.root.dummy.child2.capacity=50 > Memory Total is 20GB, default queue share is 4GB and dummy queue share is > 16GB. Child and Child1 queue gets 8GB share each. > A map reduce job is submitted to child2 queue which asks 2 containers of 512 > MB. Now cluster Memory Used is 1GB. > Root queue usage = 100 / (total memory / used memory) = 100 / (20 / 1) = 5% > Dummy queue usage = 100 / (16 /1) = 6.3% > Dummy.Child2 queue usage = 100 / (8/1) = 12.5% > At application level, % of queue is calculated as 100 / (50% of root queue > capacity) = 100 / (50% of 20GB) = 10.0 instead of > 100 / (50% of dummy queue capacity) = 100 / (50% of 16GB) = 100 / 8 = 12.5 > Where 50% is dummy.child2 capacity > Attached RM UI screenshot. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Commented] (YARN-6052) Yarn RM UI % of Queue at application level is wrong
[ https://issues.apache.org/jira/browse/YARN-6052?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15803853#comment-15803853 ] Prabhu Joseph commented on YARN-6052: - Sorry for the spam, the issue is already fixed by YARN-. Closing this as a Duplicate. > Yarn RM UI % of Queue at application level is wrong > --- > > Key: YARN-6052 > URL: https://issues.apache.org/jira/browse/YARN-6052 > Project: Hadoop YARN > Issue Type: Bug >Affects Versions: 2.7.3 >Reporter: Prabhu Joseph >Assignee: Prabhu Joseph >Priority: Minor > Attachments: RM_UI.png > > > Test Case: > yarn.scheduler.capacity.root.capacity=100 > yarn.scheduler.capacity.root.queues=default,dummy > yarn.scheduler.capacity.root.default.capacity=20 > yarn.scheduler.capacity.root.dummy.capacity=80 > yarn.scheduler.capacity.root.dummy.child.capacity=50 > yarn.scheduler.capacity.root.dummy.child2.capacity=50 > Memory Total is 20GB, default queue share is 4GB and dummy queue share is > 16GB. Child and Child1 queue gets 8GB share each. > A map reduce job is submitted to child2 queue which asks 2 containers of 512 > MB. Now cluster Memory Used is 1GB. > Root queue usage = 100 / (total memory / used memory) = 100 / (20 / 1) = 5% > Dummy queue usage = 100 / (16 /1) = 6.3% > Dummy.Child2 queue usage = 100 / (8/1) = 12.5% > At application level, % of queue is calculated as 100 / (50% of root queue > capacity) = 100 / (50% of 20GB) = 10.0 instead of > 100 / (50% of dummy queue capacity) = 100 / (50% of 16GB) = 100 / 8 = 12.5 > Where 50% is dummy.child2 capacity > Attached RM UI screenshot. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
[jira] [Commented] (YARN-6052) Yarn RM UI % of Queue at application level is wrong
[
https://issues.apache.org/jira/browse/YARN-6052?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15798205#comment-15798205
]
Sunil G commented on YARN-6052:
---
Looks like a bug. [~Prabhu Joseph], could you please provide a patch. I can
help to review. You can look {{FiCaSchedulerApp.getResourceUsageReport()}}
> Yarn RM UI % of Queue at application level is wrong
> ---
>
> Key: YARN-6052
> URL: https://issues.apache.org/jira/browse/YARN-6052
> Project: Hadoop YARN
> Issue Type: Bug
>Affects Versions: 2.7.3
>Reporter: Prabhu Joseph
>Assignee: Prabhu Joseph
>Priority: Minor
> Attachments: RM_UI.png
>
>
> Test Case:
> yarn.scheduler.capacity.root.capacity=100
> yarn.scheduler.capacity.root.queues=default,dummy
> yarn.scheduler.capacity.root.default.capacity=20
> yarn.scheduler.capacity.root.dummy.capacity=80
> yarn.scheduler.capacity.root.dummy.child.capacity=50
> yarn.scheduler.capacity.root.dummy.child2.capacity=50
> Memory Total is 20GB, default queue share is 4GB and dummy queue share is
> 16GB. Child and Child1 queue gets 8GB share each.
> A map reduce job is submitted to child2 queue which asks 2 containers of 512
> MB. Now cluster Memory Used is 1GB.
> Root queue usage = 100 / (total memory / used memory) = 100 / (20 / 1) = 5%
> Dummy queue usage = 100 / (16 /1) = 6.3%
> Dummy.Child2 queue usage = 100 / (8/1) = 12.5%
> At application level, % of queue is calculated as 100 / (50% of root queue
> capacity) = 100 / (50% of 20GB) = 10.0 instead of
> 100 / (50% of dummy queue capacity) = 100 / (50% of 16GB) = 100 / 8 = 12.5
> Where 50% is dummy.child2 capacity
> Attached RM UI screenshot.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
-
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
