Currently, grid engine autoscaler scales down workers if they have invalid states. In case a node becomes unavailable the autoscaler will scale it down immediately. This approach doesn't give node a chance to recover.
Approach
Grid engine autoscaler should be able to postpone unavailable nodes termination in the same manner as in #3297.
Background
Currently, grid engine autoscaler scales down workers if they have invalid states. In case a node becomes unavailable the autoscaler will scale it down immediately. This approach doesn't give node a chance to recover.
Approach
Grid engine autoscaler should be able to postpone unavailable nodes termination in the same manner as in #3297.