Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[k8s] Fix --purge not cleaning up cluster in stale k8s context #4514

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

[k8s] Fix --purge not cleaning up cluster in stale k8s context #4514

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e5fe41a
Fix purge not cleaning up stale k8s context cluster
romilbhardwaj Dec 29, 2024
c08fcc3
update comment
romilbhardwaj Dec 29, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
58 changes: 32 additions & 26 deletions sky/backends/cloud_vm_ray_backend.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4195,34 +4195,40 @@ def post_teardown_cleanup(self,
# that successfully call this function but do not first call
# teardown_cluster or terminate_instances. See
# https://github.com/skypilot-org/skypilot/pull/4443#discussion_r1872798032
# If purge is set, we do not care about instance status and should skip
# the check because it may fail if the cluster is not reachable.
attempts = 0
while True:
logger.debug(f'instance statuses attempt {attempts + 1}')
node_status_dict = provision_lib.query_instances(
repr(cloud),
cluster_name_on_cloud,
config['provider'],
non_terminated_only=False)

unexpected_node_state: Optional[Tuple[str, str]] = None
for node_id, node_status in node_status_dict.items():
logger.debug(f'{node_id} status: {node_status}')
# FIXME(cooperc): Some clouds (e.g. GCP) do not distinguish
# between "stopping/stopped" and "terminating/terminated", so we
# allow for either status instead of casing on `terminate`.
if node_status not in [None, status_lib.ClusterStatus.STOPPED]:
unexpected_node_state = (node_id, node_status)

if unexpected_node_state is None:
break
if not purge:
while True:
logger.debug(f'instance statuses attempt {attempts + 1}')
node_status_dict = provision_lib.query_instances(
repr(cloud),
cluster_name_on_cloud,
config['provider'],
non_terminated_only=False)

unexpected_node_state: Optional[Tuple[str, str]] = None
for node_id, node_status in node_status_dict.items():
logger.debug(f'{node_id} status: {node_status}')
# FIXME(cooperc): Some clouds (e.g. GCP) do not distinguish
# between "stopping/stopped" and "terminating/terminated",
# so we allow for either status instead of casing
# on `terminate`.
if node_status not in [
None, status_lib.ClusterStatus.STOPPED
]:
unexpected_node_state = (node_id, node_status)

if unexpected_node_state is None:
break

attempts += 1
if attempts < _TEARDOWN_WAIT_MAX_ATTEMPTS:
time.sleep(_TEARDOWN_WAIT_BETWEEN_ATTEMPS_SECONDS)
else:
(node_id, node_status) = unexpected_node_state
raise RuntimeError(f'Instance {node_id} in unexpected state '
f'{node_status}.')
attempts += 1
if attempts < _TEARDOWN_WAIT_MAX_ATTEMPTS:
time.sleep(_TEARDOWN_WAIT_BETWEEN_ATTEMPS_SECONDS)
else:
(node_id, node_status) = unexpected_node_state
raise RuntimeError(f'Instance {node_id} in unexpected '
'state {node_status}.')
Comment on lines +4230 to +4231
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Would it be possible to have the purge check here instead, so that we can correctly print out some warnings for the clusters not correctly cleaned up even when --purge is specified.


global_user_state.remove_cluster(handle.cluster_name,
terminate=terminate)
Expand Down
Loading