@ancient-wolf-19325 when marking a workflow as interruptible, how does retry happen?

When a workflow is marked as interruptible, Flyte attempts to execute tasks on spot instances, which can be preempted. If a task is preempted, Flyte retries it on a spot instance up to

n-1

times, where

n

is the number of retries specified. The last retry is executed on a non-spot instance to ensure completion. Spot instance preemptions count against the system retry budget, not the user retry budget. User retries are only used for application errors. Would you like more detail on this process? Sources: - Flyte documentation on interruptible tasks - Flyte documentation on retries 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.

@ancient-wolf-19325 So marking a workflow as interruptible but not setting retries on a task means that the task won't run on spot? But any task w/ retries > 1 will try to run on Spot?

If a task is marked as interruptible but no retries are set, it will not automatically retry on spot instances if preempted. To enable retries on spot instances, you must set at least one retry. Tasks with

retries > 1

will attempt to run on spot instances, retrying on spot instances

n-1

times, and on a non-spot instance for the last retry. Would you like more detail on configuring retries for interruptible tasks? Sources: - Flyte documentation on interruptible tasks - Flyte community discussion on retries 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.