[ 
https://issues.apache.org/jira/browse/YARN-3409?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16097025#comment-16097025
 ] 

Naganarasimha G R commented on YARN-3409:
-----------------------------------------

Thanks [[email protected]], [~grey], [~sunilg] & [~wangda] for sharing your 
view points,
We (Huawei folks) too had discussion in the same lines and in YARN we had 
already created a concept of "NodeLabels/ NodePartitions" which was not inline 
with putting everything into resource. So my view was to have it as separate in 
terms consistency and clarity. IMHO it would be more intuitive for users to 
understand it, if not clubbed as resource.

bq.  "value for the label could change between when the allocation is made and 
when the AM launches the container" ,
As [~wangda] corrected its not explicitly triggering Preemption but based on 
demand it gets triggered, but in case of attributes may be we need to rethink 
on it as the attributes might impact the ability of execution of a container. 
Assume a version of library is mapped as a attribute and containers have been 
allocated based on it , now if the library gets updated and thus the 
attribute's value. So would it be better to validate the expression of the 
containers (resource request) running on the node or let it take the natural 
course(fail and get relaunched) ? Earlier in discussion with [~sunilg] we were 
thinking of preemption based on validation of expression.

> Support Node Attribute functionality
> ------------------------------------
>
>                 Key: YARN-3409
>                 URL: https://issues.apache.org/jira/browse/YARN-3409
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: api, capacityscheduler, client
>            Reporter: Wangda Tan
>            Assignee: Naganarasimha G R
>         Attachments: Constraint-Node-Labels-Requirements-Design-doc_v1.pdf, 
> YARN-3409.WIP.001.patch
>
>
> Specify only one label for each node (IAW, partition a cluster) is a way to 
> determinate how resources of a special set of nodes could be shared by a 
> group of entities (like teams, departments, etc.). Partitions of a cluster 
> has following characteristics:
> - Cluster divided to several disjoint sub clusters.
> - ACL/priority can apply on partition (Only market team / marke team has 
> priority to use the partition).
> - Percentage of capacities can apply on partition (Market team has 40% 
> minimum capacity and Dev team has 60% of minimum capacity of the partition).
> Constraints are orthogonal to partition, they’re describing attributes of 
> node’s hardware/software just for affinity. Some example of constraints:
> - glibc version
> - JDK version
> - Type of CPU (x86_64/i686)
> - Type of OS (windows, linux, etc.)
> With this, application can be able to ask for resource has (glibc.version >= 
> 2.20 && JDK.version >= 8u20 && x86_64).



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to