Start job -> Test Job -> Clean up job

This is the order of the gitlab ci jobs in the pipeline, i have automatic retry if the start job fails, but the thing is, if the start job runs again immediately it wont work the second time as it reports that the env is exists already. This is what is happening right now.

I’d like to have some kinda cleanup (on failure of the start job) right after start job fails and then retry the start job just one more time. Tricky part that I can’t figure out is the retry of start after the on failure script.