New RunLLM response generated: >>>*# Question* Is there a way to specify only execution timeout and keep queuing timeout unlimited? the context is that we're seeing long queue times because of compute unavailability and are seeing timeouts. I see in the code that there is actually (on the propeller) active deadline + execution deadline and that setting timeout on the task sets both: https://github.com/flyteorg/flyte/blame/e3d99b591dec7d230e63bd6cb91afd120ea62432/flytepropeller/pkg/compiler/transformers/k8s/node.go#L76. Thanks for any guidance here. # Answer Currently, Flyte does not support specifying separate timeouts for execution and queuing directly through the task configuration. The

timeout

parameter in Flyte's

TaskMetadata

applies to both queuing and execution time, as indicated in the FlytePropeller code. However, there is an open feature request to introduce a

queue_timeout

flag that would allow users to specify a maximum queuing time separately from the execution timeout (GitHub Issue #1149). This feature is not yet implemented. Would you like more detail on potential workarounds or further information on the current timeout configuration? Sources: - GitHub Issue #1149 Slack Link https://.slack.com/archives/CP2HDHKE1/p1728077010.413229 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.