From 12f1208d46b3409079227f071b27f28db773b627 Mon Sep 17 00:00:00 2001
From: Chester Li <chaoleili2@gmail.com>
Date: Thu, 12 Dec 2024 11:15:24 +0800
Subject: [PATCH 01/13] [k8s] Add validation for pod_config #4206

Check pod_config when run 'sky check k8s' by using k8s api
---
 sky/provision/kubernetes/utils.py | 43 +++++++++++++++++++++++++++++++
 1 file changed, 43 insertions(+)

diff --git a/sky/provision/kubernetes/utils.py b/sky/provision/kubernetes/utils.py
index 7442c9be7a6..2c062ad987f 100644
--- a/sky/provision/kubernetes/utils.py
+++ b/sky/provision/kubernetes/utils.py
@@ -866,6 +866,15 @@ def check_credentials(context: Optional[str],
 
     _, exec_msg = is_kubeconfig_exec_auth(context)
 
+    # Check whether pod_config is valid
+    pod_config = skypilot_config.get_nested(('kubernetes', 'pod_config'),
+                                            default_value={},
+                                            override_configs={})
+    if pod_config:
+        _, pod_msg = _check_pod_config(context, pod_config)
+        if pod_msg:
+            return False, pod_msg
+
     # We now check if GPUs are available and labels are set correctly on the
     # cluster, and if not we return hints that may help debug any issues.
     # This early check avoids later surprises for user when they try to run
@@ -891,6 +900,40 @@ def check_credentials(context: Optional[str],
     else:
         return True, None
 
+def _check_pod_config(
+        context: Optional[str] = None, pod_config: Optional[Any] = None) \
+        -> Tuple[bool, Optional[str]]:
+    """Check if the pod_config is a valid pod config
+
+    Using create_namespaced_pod api with dry_run to check the pod_config
+    is valid or not.
+
+    Returns:
+        bool: True if pod_config is valid.
+        str: Error message about why the pod_config is invalid, None otherwise.
+    """
+    try:
+        namespace = get_kube_config_context_namespace(context)
+        kubernetes.core_api(context).create_namespaced_pod(
+            namespace,
+            body=pod_config,
+            dry_run='All',
+            field_validation='Strict',
+            _request_timeout=kubernetes.API_TIMEOUT)
+    except kubernetes.api_exception() as e:
+        error_msg = ''
+        if e.body:
+            # get detail error message from api_exception
+            exception_body = json.loads(e.body)
+            error_msg = exception_body.get('message')
+        else:
+            error_msg = str(e)
+        return False, f'Invalid pod_config: {error_msg}'
+    except Exception as e:  # pylint: disable=broad-except
+        return False, ('An error occurred: '
+                       f'{common_utils.format_exception(e, use_bracket=True)}')
+    return True, None
+
 
 def is_kubeconfig_exec_auth(
         context: Optional[str] = None) -> Tuple[bool, Optional[str]]:

From 64bb66ac1521270b030a21a28b5fc3c4b3503441 Mon Sep 17 00:00:00 2001
From: Chester Li <chaoleili2@gmail.com>
Date: Fri, 13 Dec 2024 16:41:44 +0800
Subject: [PATCH 02/13] update: check pod_config when launch

check merged pod_config during launch using k8s api
---
 sky/backends/backend_utils.py     |  4 ++++
 sky/provision/kubernetes/utils.py | 28 ++++++++++++----------------
 2 files changed, 16 insertions(+), 16 deletions(-)

diff --git a/sky/backends/backend_utils.py b/sky/backends/backend_utils.py
index a3651bdba9a..95c2296b0b9 100644
--- a/sky/backends/backend_utils.py
+++ b/sky/backends/backend_utils.py
@@ -916,6 +916,10 @@ def write_cluster_config(
             tmp_yaml_path,
             cluster_config_overrides=to_provision.cluster_config_overrides)
         kubernetes_utils.combine_metadata_fields(tmp_yaml_path)
+        valid, message = kubernetes_utils.check_pod_config(tmp_yaml_path)
+        if not valid:
+            raise exceptions.InvalidCloudConfigs(
+                f'There are invalid config in pod_config, deatil: {message}')
 
     if dryrun:
         # If dryrun, return the unfinished tmp yaml path.
diff --git a/sky/provision/kubernetes/utils.py b/sky/provision/kubernetes/utils.py
index 2c062ad987f..11101c2ac95 100644
--- a/sky/provision/kubernetes/utils.py
+++ b/sky/provision/kubernetes/utils.py
@@ -866,15 +866,6 @@ def check_credentials(context: Optional[str],
 
     _, exec_msg = is_kubeconfig_exec_auth(context)
 
-    # Check whether pod_config is valid
-    pod_config = skypilot_config.get_nested(('kubernetes', 'pod_config'),
-                                            default_value={},
-                                            override_configs={})
-    if pod_config:
-        _, pod_msg = _check_pod_config(context, pod_config)
-        if pod_msg:
-            return False, pod_msg
-
     # We now check if GPUs are available and labels are set correctly on the
     # cluster, and if not we return hints that may help debug any issues.
     # This early check avoids later surprises for user when they try to run
@@ -900,9 +891,8 @@ def check_credentials(context: Optional[str],
     else:
         return True, None
 
-def _check_pod_config(
-        context: Optional[str] = None, pod_config: Optional[Any] = None) \
-        -> Tuple[bool, Optional[str]]:
+
+def check_pod_config(cluster_yaml_path: str) -> Tuple[bool, Optional[str]]:
     """Check if the pod_config is a valid pod config
 
     Using create_namespaced_pod api with dry_run to check the pod_config
@@ -912,13 +902,19 @@ def _check_pod_config(
         bool: True if pod_config is valid.
         str: Error message about why the pod_config is invalid, None otherwise.
     """
+    with open(cluster_yaml_path, 'r', encoding='utf-8') as f:
+        yaml_content = f.read()
+    yaml_obj = yaml.safe_load(yaml_content)
+    pod_config = \
+        yaml_obj['available_node_types']['ray_head_default']['node_config']
     try:
-        namespace = get_kube_config_context_namespace(context)
-        kubernetes.core_api(context).create_namespaced_pod(
+        # This ok to use None context here as we only test the pod is valid
+        # won't do any change in the cluster
+        namespace = get_kube_config_context_namespace(None)
+        kubernetes.core_api().create_namespaced_pod(
             namespace,
             body=pod_config,
             dry_run='All',
-            field_validation='Strict',
             _request_timeout=kubernetes.API_TIMEOUT)
     except kubernetes.api_exception() as e:
         error_msg = ''
@@ -928,7 +924,7 @@ def _check_pod_config(
             error_msg = exception_body.get('message')
         else:
             error_msg = str(e)
-        return False, f'Invalid pod_config: {error_msg}'
+        return False, error_msg
     except Exception as e:  # pylint: disable=broad-except
         return False, ('An error occurred: '
                        f'{common_utils.format_exception(e, use_bracket=True)}')

From a0f29e5585476f49ac7db85e5d67caef0860448f Mon Sep 17 00:00:00 2001
From: Chester Li <chaoleili2@gmail.com>
Date: Fri, 13 Dec 2024 17:00:25 +0800
Subject: [PATCH 03/13] fix test

---
 sky/provision/kubernetes/utils.py | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff --git a/sky/provision/kubernetes/utils.py b/sky/provision/kubernetes/utils.py
index 11101c2ac95..495147d3caf 100644
--- a/sky/provision/kubernetes/utils.py
+++ b/sky/provision/kubernetes/utils.py
@@ -907,11 +907,11 @@ def check_pod_config(cluster_yaml_path: str) -> Tuple[bool, Optional[str]]:
     yaml_obj = yaml.safe_load(yaml_content)
     pod_config = \
         yaml_obj['available_node_types']['ray_head_default']['node_config']
+    context = yaml_obj['provider'].get('context',
+                                       get_current_kube_config_context_name())
     try:
-        # This ok to use None context here as we only test the pod is valid
-        # won't do any change in the cluster
-        namespace = get_kube_config_context_namespace(None)
-        kubernetes.core_api().create_namespaced_pod(
+        namespace = get_kube_config_context_namespace(context)
+        kubernetes.core_api(context).create_namespaced_pod(
             namespace,
             body=pod_config,
             dry_run='All',
@@ -925,6 +925,8 @@ def check_pod_config(cluster_yaml_path: str) -> Tuple[bool, Optional[str]]:
         else:
             error_msg = str(e)
         return False, error_msg
+    except ValueError as e:
+        return False, common_utils.format_exception(e)
     except Exception as e:  # pylint: disable=broad-except
         return False, ('An error occurred: '
                        f'{common_utils.format_exception(e, use_bracket=True)}')

From 47e724da33537f87c1caa8986d7a93285e7c32a5 Mon Sep 17 00:00:00 2001
From: Chester Li <chaoleili2@gmail.com>
Date: Fri, 13 Dec 2024 17:16:18 +0800
Subject: [PATCH 04/13] ignore check failed when test with dryrun

if there is no kube config in env, ignore ValueError when launch
with dryrun. For now, we don't support check schema offline.
---
 sky/backends/backend_utils.py     | 3 ++-
 sky/provision/kubernetes/utils.py | 8 +++++++-
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/sky/backends/backend_utils.py b/sky/backends/backend_utils.py
index 95c2296b0b9..62d965e3b79 100644
--- a/sky/backends/backend_utils.py
+++ b/sky/backends/backend_utils.py
@@ -916,7 +916,8 @@ def write_cluster_config(
             tmp_yaml_path,
             cluster_config_overrides=to_provision.cluster_config_overrides)
         kubernetes_utils.combine_metadata_fields(tmp_yaml_path)
-        valid, message = kubernetes_utils.check_pod_config(tmp_yaml_path)
+        valid, message = kubernetes_utils.check_pod_config(
+            tmp_yaml_path, dryrun)
         if not valid:
             raise exceptions.InvalidCloudConfigs(
                 f'There are invalid config in pod_config, deatil: {message}')
diff --git a/sky/provision/kubernetes/utils.py b/sky/provision/kubernetes/utils.py
index 495147d3caf..8cd46cf7323 100644
--- a/sky/provision/kubernetes/utils.py
+++ b/sky/provision/kubernetes/utils.py
@@ -892,7 +892,8 @@ def check_credentials(context: Optional[str],
         return True, None
 
 
-def check_pod_config(cluster_yaml_path: str) -> Tuple[bool, Optional[str]]:
+def check_pod_config(cluster_yaml_path: str, dryrun: bool) \
+    -> Tuple[bool, Optional[str]]:
     """Check if the pod_config is a valid pod config
 
     Using create_namespaced_pod api with dry_run to check the pod_config
@@ -926,6 +927,11 @@ def check_pod_config(cluster_yaml_path: str) -> Tuple[bool, Optional[str]]:
             error_msg = str(e)
         return False, error_msg
     except ValueError as e:
+        if dryrun:
+            logger.debug('ignore ValueError as there is no kube config '
+                         'in the enviroment with dry_run. '
+                         'For now we don\'t support check pod_config offline.')
+            return True, None
         return False, common_utils.format_exception(e)
     except Exception as e:  # pylint: disable=broad-except
         return False, ('An error occurred: '

From 699961d599ba561fea0ae3defc8f945686b1d388 Mon Sep 17 00:00:00 2001
From: Chester Li <chaoleili2@gmail.com>
Date: Wed, 25 Dec 2024 12:36:32 +0800
Subject: [PATCH 05/13] use deserialize api to check pod_config schema

---
 sky/backends/backend_utils.py     |  5 ++-
 sky/provision/kubernetes/utils.py | 67 +++++++++++++++----------------
 tests/test_config.py              | 54 +++++++++++++++++++++++++
 3 files changed, 89 insertions(+), 37 deletions(-)

diff --git a/sky/backends/backend_utils.py b/sky/backends/backend_utils.py
index 62d965e3b79..143876a59ab 100644
--- a/sky/backends/backend_utils.py
+++ b/sky/backends/backend_utils.py
@@ -916,8 +916,9 @@ def write_cluster_config(
             tmp_yaml_path,
             cluster_config_overrides=to_provision.cluster_config_overrides)
         kubernetes_utils.combine_metadata_fields(tmp_yaml_path)
-        valid, message = kubernetes_utils.check_pod_config(
-            tmp_yaml_path, dryrun)
+        yaml_obj = common_utils.read_yaml(tmp_yaml_path)
+        pod_config = yaml_obj['available_node_types']['ray_head_default']['node_config']
+        valid, message = kubernetes_utils.check_pod_config(pod_config)
         if not valid:
             raise exceptions.InvalidCloudConfigs(
                 f'There are invalid config in pod_config, deatil: {message}')
diff --git a/sky/provision/kubernetes/utils.py b/sky/provision/kubernetes/utils.py
index 8cd46cf7323..d0a0e87860a 100644
--- a/sky/provision/kubernetes/utils.py
+++ b/sky/provision/kubernetes/utils.py
@@ -892,51 +892,48 @@ def check_credentials(context: Optional[str],
         return True, None
 
 
-def check_pod_config(cluster_yaml_path: str, dryrun: bool) \
+def check_pod_config(pod_config: dict) \
     -> Tuple[bool, Optional[str]]:
     """Check if the pod_config is a valid pod config
 
-    Using create_namespaced_pod api with dry_run to check the pod_config
-    is valid or not.
+    Using deserialize api to check the pod_config is valid or not.
 
     Returns:
         bool: True if pod_config is valid.
         str: Error message about why the pod_config is invalid, None otherwise.
     """
-    with open(cluster_yaml_path, 'r', encoding='utf-8') as f:
-        yaml_content = f.read()
-    yaml_obj = yaml.safe_load(yaml_content)
-    pod_config = \
-        yaml_obj['available_node_types']['ray_head_default']['node_config']
-    context = yaml_obj['provider'].get('context',
-                                       get_current_kube_config_context_name())
+    errors = []
+    api_client = kubernetes.api_client()
+
+    # Used for kubernetes api_client deserialize function, the function will use
+    # data attr, the detail ref:
+    # https://github.com/kubernetes-client/python/blob/master/kubernetes/client/api_client.py#L244
+    class InnerResponse():
+
+        def __init__(self, data: dict):
+            self.data = json.dumps(data)
+
     try:
-        namespace = get_kube_config_context_namespace(context)
-        kubernetes.core_api(context).create_namespaced_pod(
-            namespace,
-            body=pod_config,
-            dry_run='All',
-            _request_timeout=kubernetes.API_TIMEOUT)
-    except kubernetes.api_exception() as e:
-        error_msg = ''
-        if e.body:
-            # get detail error message from api_exception
-            exception_body = json.loads(e.body)
-            error_msg = exception_body.get('message')
-        else:
-            error_msg = str(e)
-        return False, error_msg
-    except ValueError as e:
-        if dryrun:
-            logger.debug('ignore ValueError as there is no kube config '
-                         'in the enviroment with dry_run. '
-                         'For now we don\'t support check pod_config offline.')
-            return True, None
-        return False, common_utils.format_exception(e)
+        # Validate metadata if present
+        if 'metadata' in pod_config:
+            try:
+                value = InnerResponse(pod_config['metadata'])
+                api_client.deserialize(
+                    value, kubernetes.kubernetes.client.V1ObjectMeta)
+            except ValueError as e:
+                errors.append(f'Invalid metadata: {str(e)}')
+        # Validate spec if present
+        if 'spec' in pod_config:
+            try:
+                value = InnerResponse(pod_config['spec'])
+                api_client.deserialize(value,
+                                       kubernetes.kubernetes.client.V1PodSpec)
+            except ValueError as e:
+                errors.append(f'Invalid spec: {str(e)}')
+        return len(errors) == 0, '.'.join(errors)
     except Exception as e:  # pylint: disable=broad-except
-        return False, ('An error occurred: '
-                       f'{common_utils.format_exception(e, use_bracket=True)}')
-    return True, None
+        errors.append(f'Validation error: {str(e)}')
+        return False, '.'.join(errors)
 
 
 def is_kubeconfig_exec_auth(
diff --git a/tests/test_config.py b/tests/test_config.py
index 5789214dc61..b9f3d480d1b 100644
--- a/tests/test_config.py
+++ b/tests/test_config.py
@@ -7,6 +7,7 @@
 
 import sky
 from sky import skypilot_config
+import sky.exceptions
 from sky.skylet import constants
 from sky.utils import common_utils
 from sky.utils import kubernetes_enums
@@ -98,6 +99,39 @@ def _create_task_yaml_file(task_file_path: pathlib.Path) -> None:
         run: echo 'Running...'
         """))
 
+def _create_invalid_config_yaml_file(task_file_path: pathlib.Path) -> None:
+    task_file_path.write_text(
+        textwrap.dedent(f"""\
+        experimental:
+            config_overrides:
+                docker:
+                    run_options:
+                        - -v /tmp:/tmp
+                kubernetes:
+                    pod_config:
+                        metadata:
+                            labels:
+                                test-key: test-value
+                            annotations:
+                                abc: def
+                        spec:
+                            containers:
+                                - name:
+                            imagePullSecrets:
+                                - name: my-secret-2
+                    provision_timeout: 100
+                gcp:
+                    managed_instance_group:
+                        run_duration: {RUN_DURATION_OVERRIDE}
+                nvidia_gpus:
+                    disable_ecc: true
+        resources:
+            image_id: docker:ubuntu:latest
+
+        setup: echo 'Setting up...'
+        run: echo 'Running...'
+        """))
+
 
 def test_nested_config(monkeypatch) -> None:
     """Test that the nested config works."""
@@ -334,6 +368,26 @@ def test_k8s_config_with_override(monkeypatch, tmp_path,
                    'imagePullSecrets'][0]['name'] == 'my-secret-2'
     assert cluster_pod_config['spec']['runtimeClassName'] == 'nvidia'
 
+def test_k8s_config_with_invalid_config(monkeypatch, tmp_path,
+                                        enable_all_clouds) -> None:
+    config_path = tmp_path / 'config.yaml'
+    _create_config_file(config_path)
+    monkeypatch.setattr(skypilot_config, 'CONFIG_PATH', config_path)
+
+    _reload_config()
+    task_path = tmp_path / 'task.yaml'
+    _create_invalid_config_yaml_file(task_path)
+    task = sky.Task.from_yaml(task_path)
+
+    # Test Kubernetes pod_config invalid
+    cluster_name = 'test-kubernetes-config-with-override'
+    task.set_resources_override({'cloud': sky.Kubernetes()})
+    exception = None
+    try:
+        sky.launch(task, cluster_name=cluster_name, dryrun=True)
+    except sky.exceptions.ResourcesUnavailableError as e:
+        exception = e
+    assert not exception
 
 def test_gcp_config_with_override(monkeypatch, tmp_path,
                                   enable_all_clouds) -> None:

From 164f436d6fbae5401383c62c841441bd54bf9310 Mon Sep 17 00:00:00 2001
From: Chester Li <chaoleili2@gmail.com>
Date: Thu, 26 Dec 2024 10:04:33 +0800
Subject: [PATCH 06/13] test

---
 sky/backends/backend_utils.py | 3 ++-
 tests/test_config.py          | 5 ++++-
 2 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/sky/backends/backend_utils.py b/sky/backends/backend_utils.py
index 143876a59ab..bd22f8466ea 100644
--- a/sky/backends/backend_utils.py
+++ b/sky/backends/backend_utils.py
@@ -917,7 +917,8 @@ def write_cluster_config(
             cluster_config_overrides=to_provision.cluster_config_overrides)
         kubernetes_utils.combine_metadata_fields(tmp_yaml_path)
         yaml_obj = common_utils.read_yaml(tmp_yaml_path)
-        pod_config = yaml_obj['available_node_types']['ray_head_default']['node_config']
+        pod_config = yaml_obj['available_node_types']['ray_head_default'][
+            'node_config']
         valid, message = kubernetes_utils.check_pod_config(pod_config)
         if not valid:
             raise exceptions.InvalidCloudConfigs(
diff --git a/tests/test_config.py b/tests/test_config.py
index b9f3d480d1b..756e1fd6cf6 100644
--- a/tests/test_config.py
+++ b/tests/test_config.py
@@ -99,6 +99,7 @@ def _create_task_yaml_file(task_file_path: pathlib.Path) -> None:
         run: echo 'Running...'
         """))
 
+
 def _create_invalid_config_yaml_file(task_file_path: pathlib.Path) -> None:
     task_file_path.write_text(
         textwrap.dedent(f"""\
@@ -368,6 +369,7 @@ def test_k8s_config_with_override(monkeypatch, tmp_path,
                    'imagePullSecrets'][0]['name'] == 'my-secret-2'
     assert cluster_pod_config['spec']['runtimeClassName'] == 'nvidia'
 
+
 def test_k8s_config_with_invalid_config(monkeypatch, tmp_path,
                                         enable_all_clouds) -> None:
     config_path = tmp_path / 'config.yaml'
@@ -380,7 +382,7 @@ def test_k8s_config_with_invalid_config(monkeypatch, tmp_path,
     task = sky.Task.from_yaml(task_path)
 
     # Test Kubernetes pod_config invalid
-    cluster_name = 'test-kubernetes-config-with-override'
+    cluster_name = 'test_k8s_config_with_invalid_config'
     task.set_resources_override({'cloud': sky.Kubernetes()})
     exception = None
     try:
@@ -389,6 +391,7 @@ def test_k8s_config_with_invalid_config(monkeypatch, tmp_path,
         exception = e
     assert not exception
 
+
 def test_gcp_config_with_override(monkeypatch, tmp_path,
                                   enable_all_clouds) -> None:
     config_path = tmp_path / 'config.yaml'

From 0471b4cee0352b72919c94166fd6fd681800bab4 Mon Sep 17 00:00:00 2001
From: Chester Li <chaoleili2@gmail.com>
Date: Thu, 26 Dec 2024 10:15:59 +0800
Subject: [PATCH 07/13] create another api_client with no kubeconfig

---
 sky/provision/kubernetes/utils.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/sky/provision/kubernetes/utils.py b/sky/provision/kubernetes/utils.py
index d0a0e87860a..04337c5afc9 100644
--- a/sky/provision/kubernetes/utils.py
+++ b/sky/provision/kubernetes/utils.py
@@ -903,7 +903,9 @@ def check_pod_config(pod_config: dict) \
         str: Error message about why the pod_config is invalid, None otherwise.
     """
     errors = []
-    api_client = kubernetes.api_client()
+    # This api_client won't be used to send any requests, so there is no need to
+    # load kubeconfig
+    api_client = kubernetes.kubernetes.client.ApiClient()
 
     # Used for kubernetes api_client deserialize function, the function will use
     # data attr, the detail ref:

From 98a1d842f622bc8e17a84a30de9597e79a5f12d1 Mon Sep 17 00:00:00 2001
From: Chester Li <chaoleili2@gmail.com>
Date: Thu, 26 Dec 2024 10:30:42 +0800
Subject: [PATCH 08/13] test

---
 tests/test_config.py | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/tests/test_config.py b/tests/test_config.py
index 756e1fd6cf6..55860f3cd4d 100644
--- a/tests/test_config.py
+++ b/tests/test_config.py
@@ -384,12 +384,12 @@ def test_k8s_config_with_invalid_config(monkeypatch, tmp_path,
     # Test Kubernetes pod_config invalid
     cluster_name = 'test_k8s_config_with_invalid_config'
     task.set_resources_override({'cloud': sky.Kubernetes()})
-    exception = None
+    exception_occurred = False
     try:
         sky.launch(task, cluster_name=cluster_name, dryrun=True)
-    except sky.exceptions.ResourcesUnavailableError as e:
-        exception = e
-    assert not exception
+    except sky.exceptions.ResourcesUnavailableError:
+        exception_occurred = True
+    assert exception_occurred
 
 
 def test_gcp_config_with_override(monkeypatch, tmp_path,

From 7e501f391fb008721bbb6347808200df219a42cd Mon Sep 17 00:00:00 2001
From: Chester Li <chaoleili2@gmail.com>
Date: Thu, 26 Dec 2024 10:48:14 +0800
Subject: [PATCH 09/13] update error message

---
 sky/backends/backend_utils.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sky/backends/backend_utils.py b/sky/backends/backend_utils.py
index bd22f8466ea..072e60af411 100644
--- a/sky/backends/backend_utils.py
+++ b/sky/backends/backend_utils.py
@@ -922,7 +922,7 @@ def write_cluster_config(
         valid, message = kubernetes_utils.check_pod_config(pod_config)
         if not valid:
             raise exceptions.InvalidCloudConfigs(
-                f'There are invalid config in pod_config, deatil: {message}')
+                f'Invalid pod_config. Deatil: {message}')
 
     if dryrun:
         # If dryrun, return the unfinished tmp yaml path.

From ef562da1466b13f027c660987fef3ffd3f4912cd Mon Sep 17 00:00:00 2001
From: Chester Li <chaoleili2@gmail.com>
Date: Tue, 31 Dec 2024 10:32:55 +0800
Subject: [PATCH 10/13] update test

---
 sky/backends/backend_utils.py |  2 +-
 tests/test_config.py          | 17 ++++-------------
 2 files changed, 5 insertions(+), 14 deletions(-)

diff --git a/sky/backends/backend_utils.py b/sky/backends/backend_utils.py
index 072e60af411..74cd9fabb58 100644
--- a/sky/backends/backend_utils.py
+++ b/sky/backends/backend_utils.py
@@ -922,7 +922,7 @@ def write_cluster_config(
         valid, message = kubernetes_utils.check_pod_config(pod_config)
         if not valid:
             raise exceptions.InvalidCloudConfigs(
-                f'Invalid pod_config. Deatil: {message}')
+                f'Invalid pod_config. Deatils: {message}')
 
     if dryrun:
         # If dryrun, return the unfinished tmp yaml path.
diff --git a/tests/test_config.py b/tests/test_config.py
index 55860f3cd4d..d3d18a8dffe 100644
--- a/tests/test_config.py
+++ b/tests/test_config.py
@@ -102,12 +102,9 @@ def _create_task_yaml_file(task_file_path: pathlib.Path) -> None:
 
 def _create_invalid_config_yaml_file(task_file_path: pathlib.Path) -> None:
     task_file_path.write_text(
-        textwrap.dedent(f"""\
+        textwrap.dedent("""\
         experimental:
             config_overrides:
-                docker:
-                    run_options:
-                        - -v /tmp:/tmp
                 kubernetes:
                     pod_config:
                         metadata:
@@ -120,14 +117,6 @@ def _create_invalid_config_yaml_file(task_file_path: pathlib.Path) -> None:
                                 - name:
                             imagePullSecrets:
                                 - name: my-secret-2
-                    provision_timeout: 100
-                gcp:
-                    managed_instance_group:
-                        run_duration: {RUN_DURATION_OVERRIDE}
-                nvidia_gpus:
-                    disable_ecc: true
-        resources:
-            image_id: docker:ubuntu:latest
 
         setup: echo 'Setting up...'
         run: echo 'Running...'
@@ -387,7 +376,9 @@ def test_k8s_config_with_invalid_config(monkeypatch, tmp_path,
     exception_occurred = False
     try:
         sky.launch(task, cluster_name=cluster_name, dryrun=True)
-    except sky.exceptions.ResourcesUnavailableError:
+    except sky.exceptions.ResourcesUnavailableError as e:
+        assert 'Invalid pod_config. Deatils: Invalid spec: \
+        Invalid value for `name`, must not be `None`' in str(e)
         exception_occurred = True
     assert exception_occurred
 

From 2b5dba1eba3de434e3a589e6a90d003e28badf34 Mon Sep 17 00:00:00 2001
From: Chester Li <chaoleili2@gmail.com>
Date: Tue, 31 Dec 2024 10:42:28 +0800
Subject: [PATCH 11/13] test

---
 tests/test_config.py | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/tests/test_config.py b/tests/test_config.py
index d3d18a8dffe..566f824a3a9 100644
--- a/tests/test_config.py
+++ b/tests/test_config.py
@@ -377,8 +377,10 @@ def test_k8s_config_with_invalid_config(monkeypatch, tmp_path,
     try:
         sky.launch(task, cluster_name=cluster_name, dryrun=True)
     except sky.exceptions.ResourcesUnavailableError as e:
-        assert 'Invalid pod_config. Deatils: Invalid spec: \
-        Invalid value for `name`, must not be `None`' in str(e)
+        expect_error_message = (
+            'Invalid pod_config. Deatils: '
+            'Invalid spec: Invalid value for `name`, must not be `None`')
+        assert expect_error_message in str(e)
         exception_occurred = True
     assert exception_occurred
 

From 81029d71b7ca7b42fdb03617e53bff33267eba01 Mon Sep 17 00:00:00 2001
From: Chester Li <chaoleili2@gmail.com>
Date: Tue, 31 Dec 2024 10:48:36 +0800
Subject: [PATCH 12/13] test

---
 tests/test_config.py | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)

diff --git a/tests/test_config.py b/tests/test_config.py
index 566f824a3a9..d3eaeb261bc 100644
--- a/tests/test_config.py
+++ b/tests/test_config.py
@@ -376,11 +376,7 @@ def test_k8s_config_with_invalid_config(monkeypatch, tmp_path,
     exception_occurred = False
     try:
         sky.launch(task, cluster_name=cluster_name, dryrun=True)
-    except sky.exceptions.ResourcesUnavailableError as e:
-        expect_error_message = (
-            'Invalid pod_config. Deatils: '
-            'Invalid spec: Invalid value for `name`, must not be `None`')
-        assert expect_error_message in str(e)
+    except sky.exceptions.ResourcesUnavailableError:
         exception_occurred = True
     assert exception_occurred
 

From d349c379c9b48b7dda09f60d9291782b0691519f Mon Sep 17 00:00:00 2001
From: Romil Bhardwaj <romil.bhardwaj@gmail.com>
Date: Fri, 3 Jan 2025 09:07:38 -0800
Subject: [PATCH 13/13] Update sky/backends/backend_utils.py

---
 sky/backends/backend_utils.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sky/backends/backend_utils.py b/sky/backends/backend_utils.py
index 74cd9fabb58..07cd294787a 100644
--- a/sky/backends/backend_utils.py
+++ b/sky/backends/backend_utils.py
@@ -922,7 +922,7 @@ def write_cluster_config(
         valid, message = kubernetes_utils.check_pod_config(pod_config)
         if not valid:
             raise exceptions.InvalidCloudConfigs(
-                f'Invalid pod_config. Deatils: {message}')
+                f'Invalid pod_config. Details: {message}')
 
     if dryrun:
         # If dryrun, return the unfinished tmp yaml path.