Hi @saurabh18cs, Thanks for thinking along! I think your methodology works well but I feel like there are some downsides to it. E.g. it still launches the required compute before being told to shut down again. I read some other post that uses the time window for updating the job trigger, that might also provide a solution. By putting the window on +-23 hours we could also stop all unintended updates to trigger a run. It's a bit of a brute force method though :).