EE is always pulled on job execution, causing slow performance

Hi all!

I have AWX running on AWS EKS Fargate (latest AWX Operator version, Kubernetes 1.24). It works, but is incredibly slow, to the point where it becomes frustrating to work with. The logs of the different containers don’t really indicate any issues.

The only thing that I’ve found so far is that it takes a while (sometimes up to 5 minutes) to start the automation-job pod. For some reason it is always pulling the EE image, even if it’s set to “only pull the image if not present before running”. I saw some discussions regarding the latest tag on the default awx-ee image, so I created a custom EE image (based on the awx-ee image, but with a fixed version tag) but sadly I’m seeing the same behavior.
If I set the pull setting to “never pull container before running” the automation-job pod fails to create with the error “Container iamges xxx is not present with pull policy of Never”, even though it just pulled it 15min before for another job.

Can this have something to do with the fact that I’m running AWX on Fargate? Where is the automation-job pod looking for these images?

Thanks!

Hi all

Would anyone be able to provide some feedback on this?

Thanks in advance!
Kr

Hi,

The time between the request to start a job and the actual start of the job can be roughly divided into the following two phases.

A. From the time the job is requested until the automated job pod is created.
B. From the time the automated job pod is created to the time the job is actually started.