[
https://issues.apache.org/jira/browse/YUNIKORN-1456?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Craig Condit updated YUNIKORN-1456:
-----------------------------------
Description:
In the latest priority scheduling design, we need to dynamically determine the
priority of a queue.
There are several possible modes:
* Leaf queues
** priority.policy = default
*** Priority is defined as the maximum priority of any applications with
pending tasks + priority.offset, or MIN_INT32 (-2^32) if no applications have
pending tasks
** priority.policy = fence
*** Priority is always equal to priority.offset, or MIN_INT32 if no
applications have pending tasks
* Parent queues
** priority.policy = default
*** Priority is defined as the maximum priority of any child queues +
priority.offset
** priority.policy = fence
*** Priority is always equal to priority.offset or MIN_INT32 if no pending
tasks exist
We should expose this as a property of a queue in REST API calls as well.
was:
In the latest priority scheduling design, we need to dynamically determine the
priority of a queue.
There are several possible modes:
* Leaf queues
** priority.policy = default
*** Priority is defined as the maximum priority of any applications with
pending tasks + priority.offset, or MIN_INT32 (-2^32) if no applications have
pending tasks
** prioirty.policy = fence
*** Priority is always equal to priority.offset
* Parent queues
** priority.policy = default
*** Priority is defined as the maximum priority of any child queues +
priority.offset
** priority.policy = fence
*** Priority is always equal to priority.offset
We should expose this as a property of a queue in REST API calls as well.
> Track dynamic priority of queues
> --------------------------------
>
> Key: YUNIKORN-1456
> URL: https://issues.apache.org/jira/browse/YUNIKORN-1456
> Project: Apache YuniKorn
> Issue Type: Sub-task
> Components: core - scheduler
> Reporter: Craig Condit
> Assignee: Craig Condit
> Priority: Major
> Labels: pull-request-available
>
> In the latest priority scheduling design, we need to dynamically determine
> the priority of a queue.
> There are several possible modes:
>
> * Leaf queues
> ** priority.policy = default
> *** Priority is defined as the maximum priority of any applications with
> pending tasks + priority.offset, or MIN_INT32 (-2^32) if no applications have
> pending tasks
> ** priority.policy = fence
> *** Priority is always equal to priority.offset, or MIN_INT32 if no
> applications have pending tasks
> * Parent queues
> ** priority.policy = default
> *** Priority is defined as the maximum priority of any child queues +
> priority.offset
> ** priority.policy = fence
> *** Priority is always equal to priority.offset or MIN_INT32 if no pending
> tasks exist
> We should expose this as a property of a queue in REST API calls as well.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
