I successfully set up a karpenter deployer and set tolerations for tasks to run there, but I am still seeing jobs queued up for a long time. Rather than queue up during busy times I would like another task pod to be created which could trigger another node being created and added for the pod.
What am I missing?
in the roadmap… we laid some good ground work with termination grace period
this PR will allow you to bring your own HPA to AWX
ansible:devel
← dhageman:hpa
opened 03:35PM - 29 Dec 23 UTC
##### SUMMARY
Allows the operator to create horizontal pod autoscaling resour… ces for the web and task deployments.
It does _**not**_ provide guidance on how to configure the HPA as those details are very cluster and use case specific.
##### ISSUE TYPE
- New or Enhanced Feature
##### ADDITIONAL INFORMATION
This code does not change the default behavior of the operator. A horizontal pod autoscaler will not be created unless it is specifically configured.
The minimum configuration required is to set the option `task_max_replicas` or `web_max_replicas` to be an integer greater than `task_replicas` or `web_replicas` respecitively. This should create a HPA using the default values for a HPA which on OpenShift monitors CPU consumption.
A good configuration for HPA will require requests, limits, and health probes for the the web and task pods to be implemented.
try it out and see how it works for you more upvote on the PR gets it merged faster~
1 Like
Are there images and helm i can point to?
Is there some way this can be done on a schedule with cron or something?
som
August 29, 2024, 7:58pm
5
Did anyone ever get this working? I see the aforementioned HPA pull request was merged back in June.