This appears to only occur when I run a playbook which requires dynamic inventories to be updated. The inventories (3 sources) appear to be updated one at a time (each inventory source creates a pod for the update, the next source waits for the previous before starting it’s own pod).
I am using k9s, which I believe gets it’s metrics from the k8 metrics API. This shows, prior to playbook run, CPU 5%, memory 49%, there are no spikes when the errors occur.
I must also mention that no other AWX tasks are running at the time.
Regardless, I have increased the memory/CPU of the nodes and alas the same issue occurs.
Since we don’t have too much insight into your environment, are you able to try out Tower in another isolated k8s environment such as minikube?
If so, may we kindly ask that you do so, and let us know if the issue persists. Try running the demo OOTB project on a fresh instance of AWX. This might help us discern if the issue is within our product or an environmental issue.
Then unfortunately I can’t help you with that. The newer version of awx requires a certain version of kubernetes that is only provided by upgrading to the latest version of k3s.
The only option I have for you is to remain on awx 19.5 which is the one I think is good your version of
It’s the latest one. I am just sharing my experiences. I had a an older version of kubernetes. Then as I upgraded awx to the later versions of awx I increasingly found those errors that the op mentioned. I then read a change log mentioning this and to upgrade to a later version.
The moment I upgraded the problems I had immediately went away
I kept my awx configuration as code using redhat-cop controller configuration. Then i just wiped my awx and deployed a lesser Version. Then ran the controller configuration against my awx installation. Boom. All my awx configuration came back.