Marcus:

This is great. If we get reservations for all queues/partitions in a resource 
does it then require separate entries for each queue on the system or can a 
flag like “all” indicate it should be applied to all queues defined/available 
in the resource.

Thanks,
Sudhakar.

From: "Christie, Marcus Aaron" <[email protected]>
Reply-To: <[email protected]>
Date: Thursday, February 27, 2020 at 10:47 AM
To: "[email protected]" <[email protected]>
Subject: cluster reservations metadata

Hi Dev,

I'm working on https://issues.apache.org/jira/browse/AIRAVATA-3126 to add 
reservation metadata to the GroupComputeResourcePreference data model.  I want 
to get some feedback on the proposed data model:

* A GroupComputeResourcePreference can have one or more Reservations
  * This changes from the current data model which is that there can only be 
one reservation
* Each Reservation can have one or more queues/partitions. A cluster 
reservation applies to one or more (or all) queues of the 
GroupComputeResourcePreference's compute host.
  * This changes from the current data model in which a reservation is applied 
to all queues of a cluster, which isn't correct when a reservation has been 
created for only a subset of a compute resource's queues.
* Each Reservation has a reservationName and a startTime and endTime.

My hope is that this data model reflects how cluster reservations are used in 
practice. If I've missed anything, please let me know.

Thanks,

Marcus

Reply via email to