dagster-k8s 0.25.4rc0__py3-none-any.whl → 0.25.6__py3-none-any.whl

dagster_k8s/client.py

@@ -673,7 +673,6 @@ class DagsterKubernetesClient:
             # State checks below, see:
             # https://kubernetes.io/docs/reference/generated/kubernetes-api/v1.18/#containerstate-v1-core
             state = container_status.state
-
             if state.running is not None:
                 if wait_for_state == WaitForPodState.Ready:
                     # ready is boolean field of container status
@@ -684,6 +683,11 @@ class DagsterKubernetesClient:
                         continue
                     else:
                         ready_containers.add(container_status.name)
+                        if container_status.name in initcontainers:
+                            self.logger(
+                                f'Init container "{container_status.name}" is ready, waiting for non-init containers...'
+                            )
+                            continue
                         if initcontainers.issubset(exited_containers | ready_containers):
                             self.logger(f'Pod "{pod_name}" is ready, done waiting')
                             break

dagster_k8s/executor.py

@@ -332,7 +332,7 @@ class K8sStepHandler(StepHandler):
         container_context = self._get_container_context(step_handler_context)
 
         status = self._api_client.get_job_status(
-            namespace=container_context.namespace,
+            namespace=container_context.namespace,  # pyright: ignore[reportArgumentType]
             job_name=job_name,
         )
         if not status:

dagster_k8s/launcher.py

@@ -313,6 +313,11 @@ class K8sRunLauncher(RunLauncher, ConfigurableClass):
 
         self._launch_k8s_job_with_args(job_name, args, run)
 
+    def _get_resume_attempt_number(self, run: DagsterRun) -> Optional[int]:
+        if not self.supports_run_worker_crash_recovery:
+            return None
+        return self._instance.count_resume_run_attempts(run.run_id)
+
     def terminate(self, run_id):
         check.str_param(run_id, "run_id")
         run = self._instance.get_run_by_id(run_id)
@@ -325,7 +330,7 @@ class K8sRunLauncher(RunLauncher, ConfigurableClass):
         container_context = self.get_container_context_for_run(run)
 
         job_name = get_job_name_from_run_id(
-            run_id, resume_attempt_number=self._instance.count_resume_run_attempts(run.run_id)
+            run_id, resume_attempt_number=self._get_resume_attempt_number(run)
         )
 
         try:
@@ -367,12 +372,10 @@ class K8sRunLauncher(RunLauncher, ConfigurableClass):
         self, run: DagsterRun, include_container_logs: Optional[bool] = True
     ) -> Optional[str]:
         container_context = self.get_container_context_for_run(run)
-        if self.supports_run_worker_crash_recovery:
-            resume_attempt_number = self._instance.count_resume_run_attempts(run.run_id)
-        else:
-            resume_attempt_number = None
 
-        job_name = get_job_name_from_run_id(run.run_id, resume_attempt_number=resume_attempt_number)
+        job_name = get_job_name_from_run_id(
+            run.run_id, resume_attempt_number=self._get_resume_attempt_number(run)
+        )
         namespace = container_context.namespace
         pod_names = self._api_client.get_pod_names_in_job(job_name, namespace=namespace)
         full_msg = ""
@@ -397,7 +400,7 @@ class K8sRunLauncher(RunLauncher, ConfigurableClass):
             )
 
         else:
-            job_debug_info = self._api_client.get_job_debug_info(job_name, namespace=namespace)
+            job_debug_info = self._api_client.get_job_debug_info(job_name, namespace=namespace)  # pyright: ignore[reportArgumentType]
             full_msg = (
                 full_msg
                 + "\n\n"
@@ -411,15 +414,12 @@ class K8sRunLauncher(RunLauncher, ConfigurableClass):
     def check_run_worker_health(self, run: DagsterRun):
         container_context = self.get_container_context_for_run(run)
 
-        if self.supports_run_worker_crash_recovery:
-            resume_attempt_number = self._instance.count_resume_run_attempts(run.run_id)
-        else:
-            resume_attempt_number = None
-
-        job_name = get_job_name_from_run_id(run.run_id, resume_attempt_number=resume_attempt_number)
+        job_name = get_job_name_from_run_id(
+            run.run_id, resume_attempt_number=self._get_resume_attempt_number(run)
+        )
         try:
             status = self._api_client.get_job_status(
-                namespace=container_context.namespace,
+                namespace=container_context.namespace,  # pyright: ignore[reportArgumentType]
                 job_name=job_name,
             )
         except Exception:

dagster_k8s/models.py

@@ -69,7 +69,7 @@ def _k8s_parse_value(data: Any, classname: str, attr_name: str) -> Any:
     elif klass == object:
         return data
     elif klass == datetime.date:
-        return parse(data).date()
+        return parse(data).date()  # pyright: ignore[reportAttributeAccessIssue]
     elif klass == datetime.datetime:
         return parse(data)
     else:

dagster_k8s/ops/k8s_job_op.py

@@ -371,7 +371,10 @@ def execute_k8s_job(
             watch = kubernetes.watch.Watch()  # consider moving in to api_client
 
             api_client.wait_for_pod(
-                pod_to_watch, namespace, wait_timeout=timeout, start_time=start_time
+                pod_to_watch,
+                namespace,  # pyright: ignore[reportArgumentType]
+                wait_timeout=timeout,
+                start_time=start_time,  # pyright: ignore[reportArgumentType]
             )
 
             log_stream = watch.stream(
@@ -416,6 +419,20 @@ def execute_k8s_job(
             num_pods_to_wait_for=num_pods_to_wait_for,
         )
     except (DagsterExecutionInterruptedError, Exception) as e:
+        try:
+            pods = api_client.get_pod_names_in_job(job_name=job_name, namespace=namespace)
+            pod_debug_info = "\n\n".join(
+                [api_client.get_pod_debug_info(pod_name, namespace) for pod_name in pods]
+            )
+        except Exception:
+            context.log.exception(
+                f"Error trying to get pod debug information for failed k8s job {job_name}"
+            )
+        else:
+            context.log.error(
+                f"Debug information for failed k8s job {job_name}:\n\n{pod_debug_info}"
+            )
+
         if delete_failed_k8s_jobs:
             context.log.info(
                 f"Deleting Kubernetes job {job_name} in namespace {namespace} due to exception"

dagster_k8s/version.py

@@ -1 +1 @@
-__version__ = "0.25.4rc0"
+__version__ = "0.25.6"

{dagster_k8s-0.25.4rc0.dist-info → dagster_k8s-0.25.6.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: dagster-k8s
-Version: 0.25.4rc0
+Version: 0.25.6
 Summary: A Dagster integration for k8s
 Home-page: https://github.com/dagster-io/dagster/tree/master/python_modules/libraries/dagster-k8s
 Author: Dagster Labs
@@ -14,7 +14,7 @@ Classifier: License :: OSI Approved :: Apache Software License
 Classifier: Operating System :: OS Independent
 Requires-Python: >=3.9,<3.13
 License-File: LICENSE
-Requires-Dist: dagster ==1.9.4rc0
+Requires-Dist: dagster ==1.9.6
 Requires-Dist: kubernetes <32
 Requires-Dist: google-auth !=2.23.1
 

dagster_k8s-0.25.6.dist-info/RECORD

@@ -0,0 +1,20 @@
+dagster_k8s/__init__.py,sha256=7LyrMxxhXKdGQYClq7OJwMiVJW0KYPP-8lTDrWLwyzU,750
+dagster_k8s/client.py,sha256=t7hktemOBbS7dP7CYbfR7FqcMOQ-j1kfnRZ9Bl6El30,40044
+dagster_k8s/container_context.py,sha256=47gts4BKPrlL624rT8MMKOn8kQtY3_f2FI2nmgsDPvU,22637
+dagster_k8s/executor.py,sha256=Z8aeAHl5dMVjh_lkz0nimGAEfPaSTfmR9pLqSkGHt18,14719
+dagster_k8s/job.py,sha256=-GzUcz80cwF9NwDAv3QdE6GezSSpUzF0kLRkj1oTO34,43046
+dagster_k8s/kubernetes_version.py,sha256=jIBF12yvVweYUCmhC5AJ2Lb1JPcHxJfohJ1_hHfzS4o,38
+dagster_k8s/launcher.py,sha256=6PIlu8b9NyXAFRlw2OKhF-Mb6-QO5gL9mkXmdmlXoCw,16589
+dagster_k8s/models.py,sha256=2ZFgAhcSR06qLlMzLnEtMgZVJ8R2PKsVGu_f6K5nx3E,6317
+dagster_k8s/pipes.py,sha256=kRHLYJfZi23GJe1_TW3Z3FglEpMrMLT43UV7YpTmKr8,30439
+dagster_k8s/py.typed,sha256=la67KBlbjXN-_-DfGNcdOcjYumVpKG_Tkw-8n5dnGB4,8
+dagster_k8s/test.py,sha256=cNtcbzxytiZtd01wY5ip7KPi01y0BUQuQhohoIfAFUM,684
+dagster_k8s/utils.py,sha256=c1bHqh5f1p5RZ0JCT6WEbPPjDvbgUp3pl4nYZRaaI4s,786
+dagster_k8s/version.py,sha256=mgi0NAwnQWP7UdrUf2L9Xln12B8NoCQbqDnalCRZMpE,23
+dagster_k8s/ops/__init__.py,sha256=ur-9GrE_DRfnsFCpYan03qOY9cWbjagC8KHZFZuiCmc,113
+dagster_k8s/ops/k8s_job_op.py,sha256=ZZUz_fYXL6cxOmKJfL56wa_q6Kvh-qjYYBaYEZIMK2s,21815
+dagster_k8s-0.25.6.dist-info/LICENSE,sha256=TMatHW4_G9ldRdodEAp-l2Xa2WvsdeOh60E3v1R2jis,11349
+dagster_k8s-0.25.6.dist-info/METADATA,sha256=aq_AlQiqOR7iSnSZPV8K2p6wuwhAAhMBWy979LA6Mmw,732
+dagster_k8s-0.25.6.dist-info/WHEEL,sha256=yQN5g4mg4AybRjkgi-9yy4iQEFibGQmlz78Pik5Or-A,92
+dagster_k8s-0.25.6.dist-info/top_level.txt,sha256=wFPjskoWPlk2hOLugYCaoZhSiZdUcbCA1QZe9I4dals,12
+dagster_k8s-0.25.6.dist-info/RECORD,,

dagster_k8s-0.25.4rc0.dist-info/RECORD

@@ -1,20 +0,0 @@
-dagster_k8s/__init__.py,sha256=7LyrMxxhXKdGQYClq7OJwMiVJW0KYPP-8lTDrWLwyzU,750
-dagster_k8s/client.py,sha256=YA24A2uSmjesCAI3Y40Y8yy3-OcWs5UMvPVd8iyqRM8,39748
-dagster_k8s/container_context.py,sha256=47gts4BKPrlL624rT8MMKOn8kQtY3_f2FI2nmgsDPvU,22637
-dagster_k8s/executor.py,sha256=Py2t7VVHdL542fL5YIQLEPmFWD6f21oG0N-AUTrXfbw,14680
-dagster_k8s/job.py,sha256=-GzUcz80cwF9NwDAv3QdE6GezSSpUzF0kLRkj1oTO34,43046
-dagster_k8s/kubernetes_version.py,sha256=jIBF12yvVweYUCmhC5AJ2Lb1JPcHxJfohJ1_hHfzS4o,38
-dagster_k8s/launcher.py,sha256=Mfs4lD8pSjqC0S39d0csPJk6g-ieVuiyv680nWtSezw,16621
-dagster_k8s/models.py,sha256=OPksL8WjsdqJtGcw4WLkvsnrm1WA6vCUjqHL9dA2oYQ,6270
-dagster_k8s/pipes.py,sha256=kRHLYJfZi23GJe1_TW3Z3FglEpMrMLT43UV7YpTmKr8,30439
-dagster_k8s/py.typed,sha256=la67KBlbjXN-_-DfGNcdOcjYumVpKG_Tkw-8n5dnGB4,8
-dagster_k8s/test.py,sha256=cNtcbzxytiZtd01wY5ip7KPi01y0BUQuQhohoIfAFUM,684
-dagster_k8s/utils.py,sha256=c1bHqh5f1p5RZ0JCT6WEbPPjDvbgUp3pl4nYZRaaI4s,786
-dagster_k8s/version.py,sha256=v8qyODIT075TCv5v2_8Yh-a1u1B20HmYXWcGIfo8nRw,26
-dagster_k8s/ops/__init__.py,sha256=ur-9GrE_DRfnsFCpYan03qOY9cWbjagC8KHZFZuiCmc,113
-dagster_k8s/ops/k8s_job_op.py,sha256=uLpQE3ZJbXDiGyyRZydyp-oDQn5XvQuhRcUwd2qiJho,21124
-dagster_k8s-0.25.4rc0.dist-info/LICENSE,sha256=TMatHW4_G9ldRdodEAp-l2Xa2WvsdeOh60E3v1R2jis,11349
-dagster_k8s-0.25.4rc0.dist-info/METADATA,sha256=aovnqPbjflVxSKfLnUAcPU1bDfUFuja7UZ3xM0HLrVg,738
-dagster_k8s-0.25.4rc0.dist-info/WHEEL,sha256=yQN5g4mg4AybRjkgi-9yy4iQEFibGQmlz78Pik5Or-A,92
-dagster_k8s-0.25.4rc0.dist-info/top_level.txt,sha256=wFPjskoWPlk2hOLugYCaoZhSiZdUcbCA1QZe9I4dals,12
-dagster_k8s-0.25.4rc0.dist-info/RECORD,,