ray.serve.config.GangRuntimeFailurePolicy#

class ray.serve.config.GangRuntimeFailurePolicy(value)[source]#

Bases: str, Enum

Policy for handling runtime failures of replicas in a gang.

PublicAPI (alpha): This API is in alpha and may change before becoming stable.

RESTART_GANG = 'RESTART_GANG'#

Tear down and restart entire gang atomically when any replica fails.

RESTART_REPLICA = 'RESTART_REPLICA'#

Tear down and restart individual replica when it fails. Other replicas in the gang will continue running.