cmasopust

it's not only the inventory that times out after 5 minutes, any job that has long running tasks (no matter whether they produce output or not, e.g. a simple pause...

There is nothing in the job logs that shows any error. When running this simple playbook: `--- - hosts: localhost connection: local gather_facts: no become: no tasks: - name: Pause...

Did some further strace of the processes in the execution environment pod and found that the "init process" dumb-init is receiving the SIGTERM signal, so it looks like the pod...

looks like this may be somehow related to AKS, my AWX is also running in AKS. what kubernetes version are you running? I've now 1.22.6 and, not totally sure, we...

I've already looked at the kublet logs but I'm not sure if that message is related to our issue because this "superfluous" http response is also logged for successful jobs...

can you show me how you modified the entrypoint? I tried something similar: just connected to the pod and run the echo command there. but... it is the exit status...

@Parkhost I don't think it's related to the Kubernetes network. We first had a cluster with Azure CNI and Kubernetes version 1.21.x where no timeout happened. Then our Azure team...

@kurokobo : thanks a lot for providing detailed instructions (for someone who's at the beginning of his Kubernetes journey 😄) I've implemented your workaround and can confirm that my long...

Looks like adding "event": "FLUSH" to the json data does no longer produce the errors in awx-task and the playbook still succeeds. Having a quick look at the AWX code...

Have some bad news: although the workaround works perfectly well now for playbook runs, it breaks the inventory updates! So, only half of my problems are solved 😒