[
https://issues.apache.org/jira/browse/YUNIKORN-629?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17316844#comment-17316844
]
Wilfred Spiegelenburg commented on YUNIKORN-629:
------------------------------------------------
The core does not use any size modifiers on resources. The unit definition or
requirement for the queue config is linked to the fact that in the shim sets a
specific unit on resources passed in.
The k8shim converts memory into MB values and cpu into milli cpus. That means
the queue configs for *this* shim needs to be specified the same. If I write a
different shim that uses memory in bytes and whole cpus only the queue config
for that shim needs to follow that set of units.
The proper solution is to allow modifiers in the queue config to remove the
ambiguity. It is not a documentation issue, it is an enhancement to the config.
This conversion is already documented here:
[http://yunikorn.apache.org/docs/next/user_guide/resource_quota_management#converting-kubernetes-resources-and-quotas]
> Add size specification to queue config
> --------------------------------------
>
> Key: YUNIKORN-629
> URL: https://issues.apache.org/jira/browse/YUNIKORN-629
> Project: Apache YuniKorn
> Issue Type: New Feature
> Components: core - common
> Reporter: Weiwei Yang
> Priority: Major
>
> We need to add some doc to explain the unit YK is expecting to setup queue
> capacity, which can be added to
> http://yunikorn.apache.org/docs/next/user_guide/queue_config.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
