Hmm, interesting … what would be the use case for this?
When you do a job retry through the API, that job can only be retried once. If you want to retry the step again, you can use the new job.id returned in that PUT request and retry that.
We want to allow manual retries to give folks discretion on if failures are worrisome or not. We have soft limits internally, but that isn’t always respected and retrying a job indefinitely can be quite problematic. We’ve also recently added features that make rebuilding preferable in some ways to retrying individual jobs which we’d like to encourage. Also, we don’t want to automatically retry as to not waste compute when the jobs fail expectedly.
This would be for general workflows through the Buildkite UI for now.
Right … I’ll share your comments with our product team to analyze.