def delete(self, entity_version=None, force_delete=False):
"""
Delete the job
:param entity_version: the entity version of the job, for concurrency control.
If entity_version is provided, start will use the provided value,
and raise an exception if version is wrong.
if entity_version is not provided, start will query job runtime to
get config version and retry until version is correct.
:param force_delete: force delete a job. If set to true, it will force
a delete of the job even if it is running.The job will be first
stopped and deleted. This step cannot be undone, and the job cannot
be re-created (with same uuid) till the delete is complete.
"""
job_entity_version = (
entity_version
or self.entity_version
or self.get_status().version.value
)
while True:
request = stateless_svc.DeleteJobRequest(
job_id=v1alpha_peloton.JobID(value=self.job_id),
version=v1alpha_peloton.EntityVersion(
value=job_entity_version
),
force=force_delete,
)
try:
self.client.stateless_svc.DeleteJob(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
except grpc.RpcError as e:
# if entity version is incorrect, get entity version from job status
# and try again.
if (
e.code() == grpc.StatusCode.ABORTED
and INVALID_ENTITY_VERSION_ERR_MESSAGE in e.details()
and entity_version is None
):
job_entity_version = (
entity_version or self.get_status().version.value
)
continue
raise
break
log.info("job %s deleted", self.job_id)
def wait_for_jobmgr_available(self):
"""
utility method to wait for job manger leader to come up.
good practice to check before all write apis
"""
attempts = 0
while attempts < self.config.max_retry_attempts:
try:
request = stateless_svc.DeleteJobRequest(
job_id=v1alpha_peloton.JobID(value=self.job_id),
version=v1alpha_peloton.EntityVersion(
value="dummy-entity-version"),
)
self.client.stateless_svc.DeleteJob(
request,
metadata=self.client.jobmgr_metadata,
timeout=self.config.rpc_timeout_sec,
)
except grpc.RpcError as e:
if e.code() != grpc.StatusCode.UNAVAILABLE:
break
log.info("waiting for job manager leader")
time.sleep(self.config.sleep_time_sec)
attempts += 1