Closed
Description
WHY
Currently, there is a gap between when the RayCluster itself is ready (and cluster.wait_ready()
completes), and when the dashboard is accessible (happens a couple seconds later at most). This means that CI or scripts where job submission is performed immediately after wait_ready()
has a small chance to fail, depending on how long the dashboard takes to be ready.
WHAT
Either wait_ready has to wait for the dashboard to be ready, or we can add a second function that waits for dashboard availability. Anything that can be used in CI/scripts to block until job submission is possible.
Metadata
Metadata
Assignees
Labels
No labels
Type
Projects
Status
Done