Details
-
Bug
-
Status: Open
-
Major
-
Resolution: Unresolved
-
Slider 0.91
-
None
-
None
Description
Slider client relies on cluster wide YARN configs for RM connection tries. Max connection retry attempts is governed by yarn.resourcemanager.connect.max-wait.ms and yarn.resourcemanager.connect.retry-interval.ms configs that are configurable at cluster level. Slider client should have its own timeout/max attempts mechanism independent of YARN configs. This is to avoid scenarios where slider client waits for a long time (default 15 minutes) when RM is down.