Major Description
What is the problem?
Currently we allow multiple jobs to run against the same repository, as long as they are not the same task. James mentioned this was so that info and collect-logs could be run. There have been a couple of cases where customers with multiple backup nodes and multiple tasks have seen errors due to two tasks being run at the same time.
The error message is confusing (it talks about lockfiles) but given we control the schedule we should be able to detect these cases,.
What is the solution?
I propose we skip scheduled jobs if we find that the repo already has a running scheduled task. This means there should only ever be a single scheduled job running at any one time which will cover most of the cases where customers have run into problems.