kubernetes_scheduler: add service_account runopt so users can specify per job acls

Author: d4l3k

torchx/pipelines/kfp/adapter.py

@@ -235,6 +235,7 @@ def container_from_app(
 def resource_from_app(
     app: api.AppDef,
     queue: str,
+    service_account: Optional[str] = None,
 ) -> dsl.ResourceOp:
     """
     resource_from_app generates a KFP ResourceOp from the provided app that uses
@@ -266,5 +267,5 @@ def resource_from_app(
         action="create",
         success_condition="status.state.phase = Completed",
         failure_condition="status.state.phase = Failed",
-        k8s_resource=app_to_resource(app, queue),
+        k8s_resource=app_to_resource(app, queue, service_account=service_account),
     )

torchx/schedulers/kubernetes_scheduler.py

@@ -157,7 +157,7 @@ def sanitize_for_serialization(obj: object) -> object:
     return api.sanitize_for_serialization(obj)
 
 
-def role_to_pod(name: str, role: Role) -> "V1Pod":
+def role_to_pod(name: str, role: Role, service_account: Optional[str]) -> "V1Pod":
     from kubernetes.client.models import (  # noqa: F811 redefinition of unused
         V1Pod,
         V1PodSpec,
@@ -207,6 +207,7 @@ def role_to_pod(name: str, role: Role) -> "V1Pod":
         spec=V1PodSpec(
             containers=[container],
             restart_policy="Never",
+            service_account_name=service_account,
         ),
         metadata=V1ObjectMeta(
             annotations={
@@ -232,7 +233,9 @@ def cleanup_str(data: str) -> str:
     return "".join(re.findall(pattern, data.lower()))
 
 
-def app_to_resource(app: AppDef, queue: str) -> Dict[str, object]:
+def app_to_resource(
+    app: AppDef, queue: str, service_account: Optional[str]
+) -> Dict[str, object]:
     """
     app_to_resource creates a volcano job kubernetes resource definition from
     the provided AppDef. The resource definition can be used to launch the
@@ -263,7 +266,7 @@ def app_to_resource(app: AppDef, queue: str) -> Dict[str, object]:
             if role_idx == 0 and replica_id == 0:
                 replica_role.env["TORCHX_RANK0_HOST"] = "localhost"
 
-            pod = role_to_pod(name, replica_role)
+            pod = role_to_pod(name, replica_role, service_account)
             pod.metadata.labels.update(pod_labels(app, role_idx, role, replica_id))
             task: Dict[str, Any] = {
                 "replicas": 1,
@@ -437,7 +440,12 @@ def _submit_dryrun(
         # map any local images to the remote image
         images_to_push = self._update_app_images(app, cfg)
 
-        resource = app_to_resource(app, queue)
+        service_account = cfg.get("service_account")
+        assert service_account is None or isinstance(
+            service_account, str
+        ), "service_account must be a str"
+
+        resource = app_to_resource(app, queue, service_account)
         req = KubernetesJob(
             resource=resource,
             images_to_push=images_to_push,
@@ -470,13 +478,21 @@ def run_opts(self) -> runopts:
             default="default",
         )
         opts.add(
-            "queue", type_=str, help="Volcano queue to schedule job in", required=True
+            "queue",
+            type_=str,
+            help="Volcano queue to schedule job in",
+            required=True,
         )
         opts.add(
             "image_repo",
             type_=str,
             help="The image repository to use when pushing patched images, must have push access. Ex: example.com/your/container",
         )
+        opts.add(
+            "service_account",
+            type_=str,
+            help="The service account name to set on the pod specs",
+        )
         return opts
 
     def describe(self, app_id: str) -> Optional[DescribeAppResponse]:
@@ -540,7 +556,7 @@ def log_iter(
 
         namespace, name = app_id.split(":")
 
-        pod_name = f"{name}-{role_name}-{k}-0"
+        pod_name = cleanup_str(f"{name}-{role_name}-{k}-0")
 
         args: Dict[str, object] = {
             "name": pod_name,

torchx/schedulers/test/kubernetes_scheduler_test.py

@@ -68,7 +68,7 @@ def test_app_to_resource_resolved_macros(self) -> None:
             "torchx.schedulers.kubernetes_scheduler.make_unique"
         ) as make_unique_ctx:
             make_unique_ctx.return_value = unique_app_name
-            resource = app_to_resource(app, "test_queue")
+            resource = app_to_resource(app, "test_queue", service_account=None)
             actual_cmd = (
                 # pyre-ignore [16]
                 resource["spec"]["tasks"][0]["template"]
@@ -88,7 +88,7 @@ def test_app_to_resource_resolved_macros(self) -> None:
 
     def test_retry_policy_not_set(self) -> None:
         app = _test_app()
-        resource = app_to_resource(app, "test_queue")
+        resource = app_to_resource(app, "test_queue", service_account=None)
         self.assertListEqual(
             [
                 {"event": "PodEvicted", "action": "RestartJob"},
@@ -99,7 +99,7 @@ def test_retry_policy_not_set(self) -> None:
         )
         for role in app.roles:
             role.max_retries = 0
-        resource = app_to_resource(app, "test_queue")
+        resource = app_to_resource(app, "test_queue", service_account=None)
         self.assertFalse("policies" in resource["spec"]["tasks"][0])
         self.assertFalse("maxRetry" in resource["spec"]["tasks"][0])
 
@@ -115,7 +115,7 @@ def test_role_to_pod(self) -> None:
         )
 
         app = _test_app()
-        pod = role_to_pod("name", app.roles[0])
+        pod = role_to_pod("name", app.roles[0], service_account="srvacc")
 
         requests = {
             "cpu": "2000m",
@@ -146,6 +146,7 @@ def test_role_to_pod(self) -> None:
             spec=V1PodSpec(
                 containers=[container],
                 restart_policy="Never",
+                service_account_name="srvacc",
             ),
             metadata=V1ObjectMeta(
                 annotations={
@@ -298,6 +299,22 @@ def test_submit_dryrun_patch(self) -> None:
             },
         )
 
+    def test_submit_dryrun_service_account(self) -> None:
+        scheduler = create_scheduler("test")
+        self.assertIn("service_account", scheduler.run_opts()._opts)
+        app = _test_app()
+        cfg = {
+            "queue": "testqueue",
+            "service_account": "srvacc",
+        }
+
+        info = scheduler._submit_dryrun(app, cfg)
+        self.assertIn("'service_account_name': 'srvacc'", str(info.request.resource))
+
+        del cfg["service_account"]
+        info = scheduler._submit_dryrun(app, cfg)
+        self.assertIn("service_account_name': None", str(info.request.resource))
+
     @patch("kubernetes.client.CustomObjectsApi.create_namespaced_custom_object")
     def test_submit(self, create_namespaced_custom_object: MagicMock) -> None:
         create_namespaced_custom_object.return_value = {
@@ -426,6 +443,7 @@ def test_runopts(self) -> None:
                 "queue",
                 "namespace",
                 "image_repo",
+                "service_account",
             },
         )
 
@@ -452,7 +470,7 @@ def test_log_iter(self, read_namespaced_pod_log: MagicMock) -> None:
         read_namespaced_pod_log.return_value = "foo reg\nfoo\nbar reg\n"
         lines = scheduler.log_iter(
             app_id="testnamespace:testjob",
-            role_name="role",
+            role_name="role_blah",
             k=1,
             regex="reg",
             since=datetime.now(),
@@ -472,7 +490,7 @@ def test_log_iter(self, read_namespaced_pod_log: MagicMock) -> None:
             kwargs,
             {
                 "namespace": "testnamespace",
-                "name": "testjob-role-1-0",
+                "name": "testjob-roleblah-1-0",
                 "timestamps": True,
             },
         )