Skip to content

feat(run): zone preference for worker groups #1050

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

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

feat(run): zone preference for worker groups #1050

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
8 changes: 8 additions & 0 deletions zetta_utils/cloud_management/resource_allocation/k8s/deployment.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,6 +38,8 @@ def _get_mazepa_deployment(
gpu_accelerator_type: str | None = None,
adc_available: bool = False,
cave_secret_available: bool = False,
required_zones: list[str] | None = None,
preferred_zones: list[str] | None = None,
) -> k8s_client.V1Deployment:
name = f"run-{name}"
pod_spec = get_mazepa_pod_spec(
Expand All @@ -50,6 +52,8 @@ def _get_mazepa_deployment(
gpu_accelerator_type=gpu_accelerator_type,
adc_available=adc_available,
cave_secret_available=cave_secret_available,
required_zones=required_zones,
preferred_zones=preferred_zones,
)

pod_template = k8s_client.V1PodTemplateSpec(
Expand Down Expand Up @@ -93,6 +97,8 @@ def get_mazepa_worker_deployment( # pylint: disable=too-many-locals
gpu_accelerator_type: str | None = None,
adc_available: bool = False,
cave_secret_available: bool = False,
required_zones: list[str] | None = None,
preferred_zones: list[str] | None = None,
):
if labels is None:
labels_final = {"run_id": run_id}
Expand All @@ -117,6 +123,8 @@ def get_mazepa_worker_deployment( # pylint: disable=too-many-locals
gpu_accelerator_type=gpu_accelerator_type,
adc_available=adc_available,
cave_secret_available=cave_secret_available,
required_zones=required_zones,
preferred_zones=preferred_zones,
)


Expand Down
48 changes: 48 additions & 0 deletions zetta_utils/cloud_management/resource_allocation/k8s/pod.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,7 @@ def get_pod_spec(
image: str,
command: str,
resources: Optional[Dict[str, int | float | str]] = None,
affinity: Optional[k8s_client.V1Affinity] = None,
dns_policy: Optional[str] = "Default",
envs: Optional[List[k8s_client.V1EnvVar]] = None,
env_secret_mapping: Optional[Dict[str, str]] = None,
Expand Down Expand Up @@ -103,6 +104,7 @@ def get_pod_spec(
)

return k8s_client.V1PodSpec(
affinity=affinity,
containers=[container, sidecar_container],
dns_policy=dns_policy,
hostname=hostname,
Expand All @@ -118,6 +120,41 @@ def get_pod_spec(
)


def _get_zone_affinities(
required_zones: list[str] | None = None, preferred_zones: list[str] | None = None
):
required_zone_affinity = None
preferred_zone_affinity = None
if required_zones:
required_zone_affinity = k8s_client.V1NodeSelector(
node_selector_terms=[
k8s_client.V1NodeSelectorTerm(
match_expressions=[
k8s_client.V1NodeSelectorRequirement(
key="topology.kubernetes.io/zone",
operator="In",
values=required_zones,
)
]
)
]
)

if preferred_zones:
preferred_zone_affinity = k8s_client.V1PreferredSchedulingTerm(
weight=100,
preference=k8s_client.V1NodeSelectorTerm(
match_expressions=[
k8s_client.V1NodeSelectorRequirement(
key="topology.kubernetes.io/zone", operator="In", values=preferred_zones
)
]
),
)

return (required_zone_affinity, preferred_zone_affinity)


def get_mazepa_pod_spec(
image: str,
command: str,
Expand All @@ -129,6 +166,8 @@ def get_mazepa_pod_spec(
gpu_accelerator_type: str | None = None,
adc_available: bool = False,
cave_secret_available: bool = False,
required_zones: list[str] | None = None,
preferred_zones: list[str] | None = None,
) -> k8s_client.V1PodSpec:
schedule_toleration = k8s_client.V1Toleration(
key="worker-pool", operator="Equal", value="true", effect="NoSchedule"
Expand All @@ -146,11 +185,20 @@ def get_mazepa_pod_spec(
)
)

required_affinity, preferred_affinity = _get_zone_affinities(required_zones, preferred_zones)
affinity = k8s_client.V1Affinity(
node_affinity=k8s_client.V1NodeAffinity(
required_during_scheduling_ignored_during_execution=required_affinity,
preferred_during_scheduling_ignored_during_execution=[preferred_affinity],
)
)

return get_pod_spec(
name="zutils-worker",
image=image,
command=command,
resources=resources,
affinity=affinity,
envs=envs,
env_secret_mapping=env_secret_mapping,
node_selector=node_selector,
Expand Down
8 changes: 8 additions & 0 deletions zetta_utils/mazepa_addons/configurations/execute_on_gcp_with_sqs.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,6 +69,8 @@ class WorkerGroup:
idle_worker_timeout: int = 300
labels: dict[str, str] | None = None
gpu_accelerator_type: str | None = None
required_zones: list[str] | None = None # k8s will schedule workers in these zones
preferred_zones: list[str] | None = None # k8s will try to schedule workers in these zones


class WorkerGroupDict(TypedDict, total=False):
Expand All @@ -82,6 +84,8 @@ class WorkerGroupDict(TypedDict, total=False):
idle_worker_timeout: NotRequired[int]
labels: NotRequired[dict[str, str]]
gpu_accelerator_type: NotRequired[str]
required_zones: NotRequired[list[str]]
preferred_zones: NotRequired[list[str]]


def _get_group_taskqueue_and_contexts(
Expand Down Expand Up @@ -124,6 +128,8 @@ def _get_group_taskqueue_and_contexts(
gpu_accelerator_type=group.gpu_accelerator_type,
adc_available=adc_available,
cave_secret_available=cave_secret_available,
required_zones=group.required_zones,
preferred_zones=group.preferred_zones,
)
job_spec = k8s.get_job_spec(pod_spec=pod_spec)
scaled_job_ctx_mngr = k8s.scaled_job_ctx_mngr(
Expand Down Expand Up @@ -154,6 +160,8 @@ def _get_group_taskqueue_and_contexts(
gpu_accelerator_type=group.gpu_accelerator_type,
adc_available=adc_available,
cave_secret_available=cave_secret_available,
required_zones=group.required_zones,
preferred_zones=group.preferred_zones,
)
deployment_ctx_mngr = k8s.deployment_ctx_mngr(
execution_id,
Expand Down
Loading