# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied. See the License for the
# specific language governing permissions and limitations
# under the License.
"""Executes task in a Kubernetes POD."""
from __future__ import annotations
import datetime
import json
import logging
import math
import os
import re
import shlex
import string
from collections.abc import Container, Iterable, Mapping, Sequence
from contextlib import AbstractContextManager
from enum import Enum
from functools import cached_property
from typing import TYPE_CHECKING, Any, Callable, Literal
import kubernetes
import tenacity
from kubernetes.client import CoreV1Api, V1Pod, models as k8s
from kubernetes.client.exceptions import ApiException
from kubernetes.stream import stream
from urllib3.exceptions import HTTPError
from airflow.configuration import conf
from airflow.exceptions import (
AirflowException,
AirflowSkipException,
TaskDeferred,
)
from airflow.models import BaseOperator
from airflow.providers.cncf.kubernetes import pod_generator
from airflow.providers.cncf.kubernetes.backcompat.backwards_compat_converters import (
convert_affinity,
convert_configmap,
convert_env_vars,
convert_env_vars_or_raise_error,
convert_image_pull_secrets,
convert_pod_runtime_info_env,
convert_port,
convert_toleration,
convert_volume,
convert_volume_mount,
)
from airflow.providers.cncf.kubernetes.callbacks import ExecutionMode, KubernetesPodOperatorCallback
from airflow.providers.cncf.kubernetes.hooks.kubernetes import KubernetesHook
from airflow.providers.cncf.kubernetes.kubernetes_helper_functions import (
POD_NAME_MAX_LENGTH,
add_unique_suffix,
create_unique_id,
)
from airflow.providers.cncf.kubernetes.pod_generator import PodGenerator
from airflow.providers.cncf.kubernetes.triggers.pod import KubernetesPodTrigger
from airflow.providers.cncf.kubernetes.utils import xcom_sidecar # type: ignore[attr-defined]
from airflow.providers.cncf.kubernetes.utils.pod_manager import (
EMPTY_XCOM_RESULT,
OnFinishAction,
PodLaunchFailedException,
PodManager,
PodNotFoundException,
PodOperatorHookProtocol,
PodPhase,
container_is_succeeded,
get_container_termination_message,
)
from airflow.settings import pod_mutation_hook
from airflow.utils import yaml
from airflow.utils.helpers import prune_dict, validate_key
from airflow.version import version as airflow_version
if TYPE_CHECKING:
import jinja2
from pendulum import DateTime
from airflow.providers.cncf.kubernetes.secret import Secret
from airflow.utils.context import Context
[docs]alphanum_lower = string.ascii_lowercase + string.digits
[docs]KUBE_CONFIG_ENV_VAR = "KUBECONFIG"
[docs]class PodEventType(Enum):
"""Type of Events emitted by kubernetes pod."""
[docs]class PodReattachFailure(AirflowException):
"""When we expect to be able to find a pod but cannot."""
[docs]class PodCredentialsExpiredFailure(AirflowException):
"""When pod fails to refresh credentials."""
[docs]class KubernetesPodOperator(BaseOperator):
"""
Execute a task in a Kubernetes Pod.
.. seealso::
For more information on how to use this operator, take a look at the guide:
:ref:`howto/operator:KubernetesPodOperator`
.. note::
If you use `Google Kubernetes Engine <https://cloud.google.com/kubernetes-engine/>`__
and Airflow is not running in the same cluster, consider using
:class:`~airflow.providers.google.cloud.operators.kubernetes_engine.GKEStartPodOperator`, which
simplifies the authorization process.
:param kubernetes_conn_id: The :ref:`kubernetes connection id <howto/connection:kubernetes>`
for the Kubernetes cluster.
:param namespace: the namespace to run within kubernetes.
:param image: Docker image you wish to launch. Defaults to hub.docker.com,
but fully qualified URLS will point to custom repositories. (templated)
:param name: name of the pod in which the task will run, will be used (plus a random
suffix if random_name_suffix is True) to generate a pod id (DNS-1123 subdomain,
containing only [a-z0-9.-]).
:param random_name_suffix: if True, will generate a random suffix.
:param cmds: entrypoint of the container. (templated)
The docker images's entrypoint is used if this is not provided.
:param arguments: arguments of the entrypoint. (templated)
The docker image's CMD is used if this is not provided.
:param ports: ports for the launched pod.
:param volume_mounts: volumeMounts for the launched pod.
:param volumes: volumes for the launched pod. Includes ConfigMaps and PersistentVolumes.
:param env_vars: Environment variables initialized in the container. (templated)
:param env_from: (Optional) List of sources to populate environment variables in the container.
:param secrets: Kubernetes secrets to inject in the container.
They can be exposed as environment vars or files in a volume.
:param in_cluster: run kubernetes client with in_cluster configuration.
:param cluster_context: context that points to kubernetes cluster.
Ignored when in_cluster is True. If None, current-context is used. (templated)
:param reattach_on_restart: if the worker dies while the pod is running, reattach and monitor
during the next try. If False, always create a new pod for each try.
:param labels: labels to apply to the Pod. (templated)
:param startup_timeout_seconds: timeout in seconds to startup the pod.
:param startup_check_interval_seconds: interval in seconds to check if the pod has already started
:param get_logs: get the stdout of the base container as logs of the tasks.
:param init_container_logs: list of init containers whose logs will be published to stdout
Takes a sequence of containers, a single container name or True. If True,
all the containers logs are published.
:param container_logs: list of containers whose logs will be published to stdout
Takes a sequence of containers, a single container name or True. If True,
all the containers logs are published. Works in conjunction with get_logs param.
The default value is the base container.
:param image_pull_policy: Specify a policy to cache or always pull an image.
:param annotations: non-identifying metadata you can attach to the Pod.
Can be a large range of data, and can include characters
that are not permitted by labels. (templated)
:param container_resources: resources for the launched pod. (templated)
:param affinity: affinity scheduling rules for the launched pod.
:param config_file: The path to the Kubernetes config file. (templated)
If not specified, default value is ``~/.kube/config``
:param node_selector: A dict containing a group of scheduling rules. (templated)
:param image_pull_secrets: Any image pull secrets to be given to the pod.
If more than one secret is required, provide a
comma separated list: secret_a,secret_b
:param service_account_name: Name of the service account
:param hostnetwork: If True enable host networking on the pod.
:param host_aliases: A list of host aliases to apply to the containers in the pod.
:param tolerations: A list of kubernetes tolerations.
:param security_context: security options the pod should run with (PodSecurityContext).
:param container_security_context: security options the container should run with.
:param dnspolicy: dnspolicy for the pod.
:param dns_config: dns configuration (ip addresses, searches, options) for the pod.
:param hostname: hostname for the pod.
:param subdomain: subdomain for the pod.
:param schedulername: Specify a schedulername for the pod
:param full_pod_spec: The complete podSpec
:param init_containers: init container for the launched Pod
:param log_events_on_failure: Log the pod's events if a failure occurs
:param do_xcom_push: If True, the content of the file
/airflow/xcom/return.json in the container will also be pushed to an
XCom when the container completes.
:param pod_template_file: path to pod template file (templated)
:param pod_template_dict: pod template dictionary (templated)
:param priority_class_name: priority class name for the launched Pod
:param pod_runtime_info_envs: (Optional) A list of environment variables,
to be set in the container.
:param termination_grace_period: Termination grace period if task killed in UI,
defaults to kubernetes default
:param configmaps: (Optional) A list of names of config maps from which it collects ConfigMaps
to populate the environment variables with. The contents of the target
ConfigMap's Data field will represent the key-value pairs as environment variables.
Extends env_from.
:param skip_on_exit_code: If task exits with this exit code, leave the task
in ``skipped`` state (default: None). If set to ``None``, any non-zero
exit code will be treated as a failure.
:param base_container_name: The name of the base container in the pod. This container's logs
will appear as part of this task's logs if get_logs is True. Defaults to None. If None,
will consult the class variable BASE_CONTAINER_NAME (which defaults to "base") for the base
container name to use.
:param deferrable: Run operator in the deferrable mode.
:param poll_interval: Polling period in seconds to check for the status. Used only in deferrable mode.
:param log_pod_spec_on_failure: Log the pod's specification if a failure occurs
:param on_finish_action: What to do when the pod reaches its final state, or the execution is interrupted.
If "delete_pod", the pod will be deleted regardless its state; if "delete_succeeded_pod",
only succeeded pod will be deleted. You can set to "keep_pod" to keep the pod.
:param termination_message_policy: The termination message policy of the base container.
Default value is "File"
:param active_deadline_seconds: The active_deadline_seconds which translates to active_deadline_seconds
in V1PodSpec.
:param callbacks: KubernetesPodOperatorCallback instance contains the callbacks methods on different step
of KubernetesPodOperator.
:param logging_interval: max time in seconds that task should be in deferred state before
resuming to fetch the latest logs. If ``None``, then the task will remain in deferred state until pod
is done, and no logs will be visible until that time.
"""
# !!! Changes in KubernetesPodOperator's arguments should be also reflected in !!!
# - airflow/decorators/__init__.pyi (by a separate PR)
# This field can be overloaded at the instance level via base_container_name
[docs] BASE_CONTAINER_NAME = "base"
[docs] ISTIO_CONTAINER_NAME = "istio-proxy"
[docs] KILL_ISTIO_PROXY_SUCCESS_MSG = "HTTP/1.1 200"
[docs] POD_CHECKED_KEY = "already_checked"
[docs] POST_TERMINATION_TIMEOUT = 120
[docs] template_fields: Sequence[str] = (
"image",
"cmds",
"annotations",
"arguments",
"env_vars",
"labels",
"config_file",
"pod_template_file",
"pod_template_dict",
"namespace",
"container_resources",
"volumes",
"volume_mounts",
"cluster_context",
"env_from",
"node_selector",
"kubernetes_conn_id",
)
[docs] template_fields_renderers = {"env_vars": "py"}
def __init__(
self,
*,
kubernetes_conn_id: str | None = KubernetesHook.default_conn_name,
namespace: str | None = None,
image: str | None = None,
name: str | None = None,
random_name_suffix: bool = True,
cmds: list[str] | None = None,
arguments: list[str] | None = None,
ports: list[k8s.V1ContainerPort] | None = None,
volume_mounts: list[k8s.V1VolumeMount] | None = None,
volumes: list[k8s.V1Volume] | None = None,
env_vars: list[k8s.V1EnvVar] | dict[str, str] | None = None,
env_from: list[k8s.V1EnvFromSource] | None = None,
secrets: list[Secret] | None = None,
in_cluster: bool | None = None,
cluster_context: str | None = None,
labels: dict | None = None,
reattach_on_restart: bool = True,
startup_timeout_seconds: int = 120,
startup_check_interval_seconds: int = 5,
get_logs: bool = True,
base_container_name: str | None = None,
init_container_logs: Iterable[str] | str | Literal[True] | None = None,
container_logs: Iterable[str] | str | Literal[True] | None = None,
image_pull_policy: str | None = None,
annotations: dict | None = None,
container_resources: k8s.V1ResourceRequirements | None = None,
affinity: k8s.V1Affinity | None = None,
config_file: str | None = None,
node_selector: dict | None = None,
image_pull_secrets: list[k8s.V1LocalObjectReference] | None = None,
service_account_name: str | None = None,
hostnetwork: bool = False,
host_aliases: list[k8s.V1HostAlias] | None = None,
tolerations: list[k8s.V1Toleration] | None = None,
security_context: k8s.V1PodSecurityContext | dict | None = None,
container_security_context: k8s.V1SecurityContext | dict | None = None,
dnspolicy: str | None = None,
dns_config: k8s.V1PodDNSConfig | None = None,
hostname: str | None = None,
subdomain: str | None = None,
schedulername: str | None = None,
full_pod_spec: k8s.V1Pod | None = None,
init_containers: list[k8s.V1Container] | None = None,
log_events_on_failure: bool = False,
do_xcom_push: bool = False,
pod_template_file: str | None = None,
pod_template_dict: dict | None = None,
priority_class_name: str | None = None,
pod_runtime_info_envs: list[k8s.V1EnvVar] | None = None,
termination_grace_period: int | None = None,
configmaps: list[str] | None = None,
skip_on_exit_code: int | Container[int] | None = None,
deferrable: bool = conf.getboolean("operators", "default_deferrable", fallback=False),
poll_interval: float = 2,
log_pod_spec_on_failure: bool = True,
on_finish_action: str = "delete_pod",
is_delete_operator_pod: None | bool = None,
termination_message_policy: str = "File",
active_deadline_seconds: int | None = None,
callbacks: type[KubernetesPodOperatorCallback] | None = None,
progress_callback: Callable[[str], None] | None = None,
logging_interval: int | None = None,
**kwargs,
) -> None:
super().__init__(**kwargs)
self.kubernetes_conn_id = kubernetes_conn_id
self.do_xcom_push = do_xcom_push
self.image = image
self.namespace = namespace
self.cmds = cmds or []
self.arguments = arguments or []
self.labels = labels or {}
self.startup_timeout_seconds = startup_timeout_seconds
self.startup_check_interval_seconds = startup_check_interval_seconds
env_vars = convert_env_vars(env_vars) if env_vars else []
self.env_vars = env_vars
pod_runtime_info_envs = (
[convert_pod_runtime_info_env(p) for p in pod_runtime_info_envs] if pod_runtime_info_envs else []
)
self.pod_runtime_info_envs = pod_runtime_info_envs
self.env_from = env_from or []
if configmaps:
self.env_from.extend([convert_configmap(c) for c in configmaps])
self.ports = [convert_port(p) for p in ports] if ports else []
volume_mounts = [convert_volume_mount(v) for v in volume_mounts] if volume_mounts else []
self.volume_mounts = volume_mounts
volumes = [convert_volume(volume) for volume in volumes] if volumes else []
self.volumes = volumes
self.secrets = secrets or []
self.in_cluster = in_cluster
self.cluster_context = cluster_context
self.reattach_on_restart = reattach_on_restart
self.get_logs = get_logs
# Fallback to the class variable BASE_CONTAINER_NAME here instead of via default argument value
# in the init method signature, to be compatible with subclasses overloading the class variable value.
self.base_container_name = base_container_name or self.BASE_CONTAINER_NAME
self.init_container_logs = init_container_logs
self.container_logs = container_logs or self.base_container_name
self.image_pull_policy = image_pull_policy
self.node_selector = node_selector or {}
self.annotations = annotations or {}
self.affinity = convert_affinity(affinity) if affinity else {}
self.container_resources = container_resources
self.config_file = config_file
self.image_pull_secrets = convert_image_pull_secrets(image_pull_secrets) if image_pull_secrets else []
self.service_account_name = service_account_name
self.hostnetwork = hostnetwork
self.host_aliases = host_aliases
self.tolerations = (
[convert_toleration(toleration) for toleration in tolerations] if tolerations else []
)
self.security_context = security_context or {}
self.container_security_context = container_security_context
self.dnspolicy = dnspolicy
self.dns_config = dns_config
self.hostname = hostname
self.subdomain = subdomain
self.schedulername = schedulername
self.full_pod_spec = full_pod_spec
self.init_containers = init_containers or []
self.log_events_on_failure = log_events_on_failure
self.priority_class_name = priority_class_name
self.pod_template_file = pod_template_file
self.pod_template_dict = pod_template_dict
self.name = self._set_name(name)
self.random_name_suffix = random_name_suffix
self.termination_grace_period = termination_grace_period
self.pod_request_obj: k8s.V1Pod | None = None
self.pod: k8s.V1Pod | None = None
self.skip_on_exit_code = (
skip_on_exit_code
if isinstance(skip_on_exit_code, Container)
else [skip_on_exit_code]
if skip_on_exit_code is not None
else []
)
self.deferrable = deferrable
self.poll_interval = poll_interval
self.remote_pod: k8s.V1Pod | None = None
self.log_pod_spec_on_failure = log_pod_spec_on_failure
self.on_finish_action = OnFinishAction(on_finish_action)
# The `is_delete_operator_pod` parameter should have been removed in provider version 10.0.0.
# TODO: remove it from here and from the operator's parameters list when the next major version bumped
self._is_delete_operator_pod = self.on_finish_action == OnFinishAction.DELETE_POD
self.termination_message_policy = termination_message_policy
self.active_deadline_seconds = active_deadline_seconds
self.logging_interval = logging_interval
self._config_dict: dict | None = None # TODO: remove it when removing convert_config_file_to_dict
self._progress_callback = progress_callback
self.callbacks = callbacks
self._killed: bool = False
@cached_property
def _incluster_namespace(self):
from pathlib import Path
path = Path("/var/run/secrets/kubernetes.io/serviceaccount/namespace")
return path.exists() and path.read_text() or None
def _render_nested_template_fields(
self,
content: Any,
context: Mapping[str, Any],
jinja_env: jinja2.Environment,
seen_oids: set,
) -> None:
if id(content) not in seen_oids:
template_fields: tuple | None
if isinstance(content, k8s.V1EnvVar):
template_fields = ("value", "name")
elif isinstance(content, k8s.V1ResourceRequirements):
template_fields = ("limits", "requests")
elif isinstance(content, k8s.V1Volume):
template_fields = ("name", "persistent_volume_claim", "config_map")
elif isinstance(content, k8s.V1VolumeMount):
template_fields = ("name", "sub_path")
elif isinstance(content, k8s.V1PersistentVolumeClaimVolumeSource):
template_fields = ("claim_name",)
elif isinstance(content, k8s.V1ConfigMapVolumeSource):
template_fields = ("name",)
elif isinstance(content, k8s.V1EnvFromSource):
template_fields = ("config_map_ref",)
elif isinstance(content, k8s.V1ConfigMapEnvSource):
template_fields = ("name",)
else:
template_fields = None
if template_fields:
seen_oids.add(id(content))
self._do_render_template_fields(content, template_fields, context, jinja_env, seen_oids)
return
super()._render_nested_template_fields(content, context, jinja_env, seen_oids)
@staticmethod
def _get_ti_pod_labels(context: Context | None = None, include_try_number: bool = True) -> dict[str, str]:
"""
Generate labels for the pod to track the pod in case of Operator crash.
:param context: task context provided by airflow DAG.
:param include_try_number: if set to True will add the try number
from the task context to the pod labels.
:return: dict
"""
if not context:
return {}
ti = context["ti"]
run_id = context["run_id"]
labels = {
"dag_id": ti.dag_id,
"task_id": ti.task_id,
"run_id": run_id,
"kubernetes_pod_operator": "True",
}
map_index = ti.map_index
if map_index >= 0:
labels["map_index"] = map_index
if include_try_number:
labels.update(try_number=ti.try_number)
# In the case of sub dags this is just useful
# TODO: Remove this when the minimum version of Airflow is bumped to 3.0
if getattr(context["dag"], "parent_dag", False):
labels["parent_dag_id"] = context["dag"].parent_dag.dag_id # type: ignore[attr-defined]
# Ensure that label is valid for Kube,
# and if not truncate/remove invalid chars and replace with short hash.
for label_id, label in labels.items():
safe_label = pod_generator.make_safe_label_value(str(label))
labels[label_id] = safe_label
return labels
@cached_property
[docs] def pod_manager(self) -> PodManager:
return PodManager(kube_client=self.client, callbacks=self.callbacks)
@cached_property
[docs] def hook(self) -> PodOperatorHookProtocol:
hook = KubernetesHook(
conn_id=self.kubernetes_conn_id,
in_cluster=self.in_cluster,
config_file=self.config_file,
cluster_context=self.cluster_context,
)
return hook
@cached_property
[docs] def client(self) -> CoreV1Api:
client = self.hook.core_v1_client
if self.callbacks:
self.callbacks.on_sync_client_creation(client=client)
return client
[docs] def find_pod(self, namespace: str, context: Context, *, exclude_checked: bool = True) -> k8s.V1Pod | None:
"""Return an already-running pod for this task instance if one exists."""
label_selector = self._build_find_pod_label_selector(context, exclude_checked=exclude_checked)
pod_list = self.client.list_namespaced_pod(
namespace=namespace,
label_selector=label_selector,
).items
pod = None
num_pods = len(pod_list)
if num_pods == 1:
pod = pod_list[0]
self.log_matching_pod(pod=pod, context=context)
elif num_pods > 1:
if self.reattach_on_restart:
raise AirflowException(f"More than one pod running with labels {label_selector}")
self.log.warning("Found more than one pod running with labels %s, resolving ...", label_selector)
pod = self.process_duplicate_label_pods(pod_list)
self.log_matching_pod(pod=pod, context=context)
return pod
[docs] def log_matching_pod(self, pod: k8s.V1Pod, context: Context) -> None:
self.log.info("Found matching pod %s with labels %s", pod.metadata.name, pod.metadata.labels)
self.log.info("`try_number` of task_instance: %s", context["ti"].try_number)
self.log.info("`try_number` of pod: %s", pod.metadata.labels["try_number"])
[docs] def get_or_create_pod(self, pod_request_obj: k8s.V1Pod, context: Context) -> k8s.V1Pod:
if self.reattach_on_restart:
pod = self.find_pod(pod_request_obj.metadata.namespace, context=context)
if pod:
return pod
self.log.debug("Starting pod:\n%s", yaml.safe_dump(pod_request_obj.to_dict()))
self.pod_manager.create_pod(pod=pod_request_obj)
return pod_request_obj
[docs] def await_pod_start(self, pod: k8s.V1Pod) -> None:
try:
self.pod_manager.await_pod_start(
pod=pod,
startup_timeout=self.startup_timeout_seconds,
startup_check_interval=self.startup_check_interval_seconds,
)
except PodLaunchFailedException:
if self.log_events_on_failure:
self._read_pod_events(pod, reraise=False)
raise
[docs] def execute(self, context: Context):
"""Based on the deferrable parameter runs the pod asynchronously or synchronously."""
if not self.deferrable:
return self.execute_sync(context)
self.execute_async(context)
[docs] def execute_sync(self, context: Context):
result = None
try:
if self.pod_request_obj is None:
self.pod_request_obj = self.build_pod_request_obj(context)
if self.pod is None:
self.pod = self.get_or_create_pod( # must set `self.pod` for `on_kill`
pod_request_obj=self.pod_request_obj,
context=context,
)
# push to xcom now so that if there is an error we still have the values
ti = context["ti"]
ti.xcom_push(key="pod_name", value=self.pod.metadata.name)
ti.xcom_push(key="pod_namespace", value=self.pod.metadata.namespace)
# get remote pod for use in cleanup methods
self.remote_pod = self.find_pod(self.pod.metadata.namespace, context=context)
if self.callbacks:
self.callbacks.on_pod_creation(
pod=self.remote_pod, client=self.client, mode=ExecutionMode.SYNC
)
self.await_init_containers_completion(pod=self.pod)
self.await_pod_start(pod=self.pod)
if self.callbacks:
self.callbacks.on_pod_starting(
pod=self.find_pod(self.pod.metadata.namespace, context=context),
client=self.client,
mode=ExecutionMode.SYNC,
)
self.await_pod_completion(pod=self.pod)
if self.callbacks:
self.callbacks.on_pod_completion(
pod=self.find_pod(self.pod.metadata.namespace, context=context),
client=self.client,
mode=ExecutionMode.SYNC,
)
if self.do_xcom_push:
self.pod_manager.await_xcom_sidecar_container_start(pod=self.pod)
result = self.extract_xcom(pod=self.pod)
istio_enabled = self.is_istio_enabled(self.pod)
self.remote_pod = self.pod_manager.await_pod_completion(
self.pod, istio_enabled, self.base_container_name
)
finally:
pod_to_clean = self.pod or self.pod_request_obj
self.cleanup(
pod=pod_to_clean,
remote_pod=self.remote_pod,
)
if self.callbacks:
self.callbacks.on_pod_cleanup(pod=pod_to_clean, client=self.client, mode=ExecutionMode.SYNC)
if self.do_xcom_push:
return result
@tenacity.retry(
wait=tenacity.wait_exponential(max=15),
retry=tenacity.retry_if_exception_type(PodCredentialsExpiredFailure),
reraise=True,
)
[docs] def await_init_containers_completion(self, pod: k8s.V1Pod):
try:
if self.init_container_logs:
self.pod_manager.fetch_requested_init_container_logs(
pod=pod,
init_containers=self.init_container_logs,
follow_logs=True,
)
except kubernetes.client.exceptions.ApiException as exc:
self._handle_api_exception(exc, pod)
@tenacity.retry(
wait=tenacity.wait_exponential(max=15),
retry=tenacity.retry_if_exception_type(PodCredentialsExpiredFailure),
reraise=True,
)
[docs] def await_pod_completion(self, pod: k8s.V1Pod):
try:
if self.get_logs:
self.pod_manager.fetch_requested_container_logs(
pod=pod,
containers=self.container_logs,
follow_logs=True,
)
if not self.get_logs or (
self.container_logs is not True and self.base_container_name not in self.container_logs
):
self.pod_manager.await_container_completion(pod=pod, container_name=self.base_container_name)
except kubernetes.client.exceptions.ApiException as exc:
self._handle_api_exception(exc, pod)
def _handle_api_exception(
self,
exc: kubernetes.client.exceptions.ApiException,
pod: k8s.V1Pod,
):
if exc.status and str(exc.status) == "401":
self.log.warning(
"Failed to check container status due to permission error. Refreshing credentials and retrying."
)
self._refresh_cached_properties()
self.pod_manager.read_pod(pod=pod) # attempt using refreshed credentials, raises if still invalid
raise PodCredentialsExpiredFailure("Kubernetes credentials expired, retrying after refresh.")
raise exc
def _refresh_cached_properties(self):
del self.hook
del self.client
del self.pod_manager
[docs] def execute_async(self, context: Context) -> None:
self.pod_request_obj = self.build_pod_request_obj(context)
self.pod = self.get_or_create_pod( # must set `self.pod` for `on_kill`
pod_request_obj=self.pod_request_obj,
context=context,
)
if self.callbacks:
self.callbacks.on_pod_creation(
pod=self.find_pod(self.pod.metadata.namespace, context=context),
client=self.client,
mode=ExecutionMode.SYNC,
)
ti = context["ti"]
ti.xcom_push(key="pod_name", value=self.pod.metadata.name)
ti.xcom_push(key="pod_namespace", value=self.pod.metadata.namespace)
self.convert_config_file_to_dict()
self.invoke_defer_method()
[docs] def convert_config_file_to_dict(self):
"""Convert passed config_file to dict representation."""
config_file = self.config_file if self.config_file else os.environ.get(KUBE_CONFIG_ENV_VAR)
if config_file:
with open(config_file) as f:
self._config_dict = yaml.safe_load(f)
else:
self._config_dict = None
[docs] def invoke_defer_method(self, last_log_time: DateTime | None = None) -> None:
"""Redefine triggers which are being used in child classes."""
trigger_start_time = datetime.datetime.now(tz=datetime.timezone.utc)
self.defer(
trigger=KubernetesPodTrigger(
pod_name=self.pod.metadata.name, # type: ignore[union-attr]
pod_namespace=self.pod.metadata.namespace, # type: ignore[union-attr]
trigger_start_time=trigger_start_time,
kubernetes_conn_id=self.kubernetes_conn_id,
cluster_context=self.cluster_context,
config_dict=self._config_dict,
in_cluster=self.in_cluster,
poll_interval=self.poll_interval,
get_logs=self.get_logs,
startup_timeout=self.startup_timeout_seconds,
startup_check_interval=self.startup_check_interval_seconds,
base_container_name=self.base_container_name,
on_finish_action=self.on_finish_action.value,
last_log_time=last_log_time,
logging_interval=self.logging_interval,
),
method_name="trigger_reentry",
)
[docs] def trigger_reentry(self, context: Context, event: dict[str, Any]) -> Any:
"""
Point of re-entry from trigger.
If ``logging_interval`` is None, then at this point, the pod should be done, and we'll just fetch
the logs and exit.
If ``logging_interval`` is not None, it could be that the pod is still running, and we'll just
grab the latest logs and defer back to the trigger again.
"""
self.pod = None
try:
pod_name = event["name"]
pod_namespace = event["namespace"]
self.pod = self.hook.get_pod(pod_name, pod_namespace)
if not self.pod:
raise PodNotFoundException("Could not find pod after resuming from deferral")
if self.callbacks and event["status"] != "running":
self.callbacks.on_operator_resuming(
pod=self.pod, event=event, client=self.client, mode=ExecutionMode.SYNC
)
follow = self.logging_interval is None
last_log_time = event.get("last_log_time")
if event["status"] in ("error", "failed", "timeout"):
# fetch some logs when pod is failed
if self.get_logs:
self._write_logs(self.pod, follow=follow, since_time=last_log_time)
if self.do_xcom_push:
_ = self.extract_xcom(pod=self.pod)
message = event.get("stack_trace", event["message"])
raise AirflowException(message)
elif event["status"] == "running":
if self.get_logs:
self.log.info("Resuming logs read from time %r", last_log_time)
pod_log_status = self.pod_manager.fetch_container_logs(
pod=self.pod,
container_name=self.base_container_name,
follow=follow,
since_time=last_log_time,
)
self.invoke_defer_method(pod_log_status.last_log_time)
else:
self.invoke_defer_method()
elif event["status"] == "success":
# fetch some logs when pod is executed successfully
if self.get_logs:
self._write_logs(self.pod, follow=follow, since_time=last_log_time)
if self.do_xcom_push:
xcom_sidecar_output = self.extract_xcom(pod=self.pod)
return xcom_sidecar_output
return
except TaskDeferred:
raise
finally:
self._clean(event)
def _clean(self, event: dict[str, Any]) -> None:
if event["status"] == "running":
return
istio_enabled = self.is_istio_enabled(self.pod)
# Skip await_pod_completion when the event is 'timeout' due to the pod can hang
# on the ErrImagePull or ContainerCreating step and it will never complete
if event["status"] != "timeout":
try:
self.pod = self.pod_manager.await_pod_completion(
self.pod, istio_enabled, self.base_container_name
)
except ApiException as e:
if e.status == 404:
self.pod = None
self.log.warning(
"Pod not found while waiting for completion. The last status was %r", event["status"]
)
else:
raise e
if self.pod is not None:
self.post_complete_action(
pod=self.pod,
remote_pod=self.pod,
)
def _write_logs(self, pod: k8s.V1Pod, follow: bool = False, since_time: DateTime | None = None) -> None:
try:
since_seconds = (
math.ceil((datetime.datetime.now(tz=datetime.timezone.utc) - since_time).total_seconds())
if since_time
else None
)
logs = self.client.read_namespaced_pod_log(
name=pod.metadata.name,
namespace=pod.metadata.namespace,
container=self.base_container_name,
follow=follow,
timestamps=False,
since_seconds=since_seconds,
_preload_content=False,
)
for raw_line in logs:
line = raw_line.decode("utf-8", errors="backslashreplace").rstrip("\n")
if line:
self.log.info("[%s] logs: %s", self.base_container_name, line)
except (HTTPError, ApiException) as e:
self.log.warning(
"Reading of logs interrupted with error %r; will retry. "
"Set log level to DEBUG for traceback.",
e if not isinstance(e, ApiException) else e.reason,
)
[docs] def post_complete_action(self, *, pod, remote_pod, **kwargs) -> None:
"""Actions that must be done after operator finishes logic of the deferrable_execution."""
self.cleanup(
pod=pod,
remote_pod=remote_pod,
)
if self.callbacks:
self.callbacks.on_pod_cleanup(pod=pod, client=self.client, mode=ExecutionMode.SYNC)
[docs] def cleanup(self, pod: k8s.V1Pod, remote_pod: k8s.V1Pod):
# Skip cleaning the pod in the following scenarios.
# 1. If a task got marked as failed, "on_kill" method would be called and the pod will be cleaned up
# there. Cleaning it up again will raise an exception (which might cause retry).
# 2. remote pod is null (ex: pod creation failed)
if self._killed or not remote_pod:
return
istio_enabled = self.is_istio_enabled(remote_pod)
pod_phase = remote_pod.status.phase if hasattr(remote_pod, "status") else None
# if the pod fails or success, but we don't want to delete it
if pod_phase != PodPhase.SUCCEEDED or self.on_finish_action == OnFinishAction.KEEP_POD:
self.patch_already_checked(remote_pod, reraise=False)
failed = (pod_phase != PodPhase.SUCCEEDED and not istio_enabled) or (
istio_enabled and not container_is_succeeded(remote_pod, self.base_container_name)
)
if failed:
if self.log_events_on_failure:
self._read_pod_events(pod, reraise=False)
self.process_pod_deletion(remote_pod, reraise=False)
if self.skip_on_exit_code:
container_statuses = (
remote_pod.status.container_statuses if remote_pod and remote_pod.status else None
) or []
base_container_status = next(
(x for x in container_statuses if x.name == self.base_container_name), None
)
exit_code = (
base_container_status.state.terminated.exit_code
if base_container_status
and base_container_status.state
and base_container_status.state.terminated
else None
)
if exit_code in self.skip_on_exit_code:
raise AirflowSkipException(
f"Pod {pod and pod.metadata.name} returned exit code {exit_code}. Skipping."
)
if failed:
error_message = get_container_termination_message(remote_pod, self.base_container_name)
raise AirflowException(
"\n".join(
filter(
None,
[
f"Pod {pod and pod.metadata.name} returned a failure.",
error_message if isinstance(error_message, str) else None,
f"remote_pod: {remote_pod}" if self.log_pod_spec_on_failure else None,
],
)
)
)
def _read_pod_events(self, pod, *, reraise=True) -> None:
"""Will fetch and emit events from pod."""
with _optionally_suppress(reraise=reraise):
for event in self.pod_manager.read_pod_events(pod).items:
if event.type == PodEventType.NORMAL.value:
self.log.info("Pod Event: %s - %s", event.reason, event.message)
else:
self.log.error("Pod Event: %s - %s", event.reason, event.message)
[docs] def is_istio_enabled(self, pod: V1Pod) -> bool:
"""Check if istio is enabled for the namespace of the pod by inspecting the namespace labels."""
if not pod:
return False
remote_pod = self.pod_manager.read_pod(pod)
return any(container.name == self.ISTIO_CONTAINER_NAME for container in remote_pod.spec.containers)
[docs] def kill_istio_sidecar(self, pod: V1Pod) -> None:
command = "/bin/sh -c 'curl -fsI -X POST http://localhost:15020/quitquitquit'"
command_to_container = shlex.split(command)
resp = stream(
self.client.connect_get_namespaced_pod_exec,
name=pod.metadata.name,
namespace=pod.metadata.namespace,
container=self.ISTIO_CONTAINER_NAME,
command=command_to_container,
stderr=True,
stdin=True,
stdout=True,
tty=False,
_preload_content=False,
)
output = []
while resp.is_open():
if resp.peek_stdout():
output.append(resp.read_stdout())
resp.close()
output_str = "".join(output)
self.log.info("Output of curl command to kill istio: %s", output_str)
resp.close()
if self.KILL_ISTIO_PROXY_SUCCESS_MSG not in output_str:
raise AirflowException("Error while deleting istio-proxy sidecar: %s", output_str)
[docs] def process_pod_deletion(self, pod: k8s.V1Pod, *, reraise=True):
with _optionally_suppress(reraise=reraise):
if pod is not None:
should_delete_pod = (self.on_finish_action == OnFinishAction.DELETE_POD) or (
self.on_finish_action == OnFinishAction.DELETE_SUCCEEDED_POD
and (
pod.status.phase == PodPhase.SUCCEEDED
or container_is_succeeded(pod, self.base_container_name)
)
)
if should_delete_pod:
self.log.info("Deleting pod: %s", pod.metadata.name)
self.pod_manager.delete_pod(pod)
else:
self.log.info("Skipping deleting pod: %s", pod.metadata.name)
def _build_find_pod_label_selector(self, context: Context | None = None, *, exclude_checked=True) -> str:
labels = {
**self.labels,
**self._get_ti_pod_labels(context, include_try_number=False),
}
label_strings = [f"{label_id}={label}" for label_id, label in sorted(labels.items())]
labels_value = ",".join(label_strings)
if exclude_checked:
labels_value = f"{labels_value},{self.POD_CHECKED_KEY}!=True"
labels_value = f"{labels_value},!airflow-worker"
return labels_value
@staticmethod
def _set_name(name: str | None) -> str | None:
if name is not None:
validate_key(name, max_length=220)
return re.sub(r"[^a-z0-9-]+", "-", name.lower())
return None
[docs] def patch_already_checked(self, pod: k8s.V1Pod, *, reraise=True):
"""Add an "already checked" label to ensure we don't reattach on retries."""
with _optionally_suppress(reraise=reraise):
self.client.patch_namespaced_pod(
name=pod.metadata.name,
namespace=pod.metadata.namespace,
body={"metadata": {"labels": {self.POD_CHECKED_KEY: "True"}}},
)
[docs] def on_kill(self) -> None:
self._killed = True
if self.pod:
pod = self.pod
kwargs = {
"name": pod.metadata.name,
"namespace": pod.metadata.namespace,
}
if self.termination_grace_period is not None:
kwargs.update(grace_period_seconds=self.termination_grace_period)
try:
self.client.delete_namespaced_pod(**kwargs)
except kubernetes.client.exceptions.ApiException:
self.log.exception("Unable to delete pod %s", self.pod.metadata.name)
[docs] def build_pod_request_obj(self, context: Context | None = None) -> k8s.V1Pod:
"""
Return V1Pod object based on pod template file, full pod spec, and other operator parameters.
The V1Pod attributes are derived (in order of precedence) from operator params, full pod spec, pod
template file.
"""
self.log.debug("Creating pod for KubernetesPodOperator task %s", self.task_id)
self.env_vars = convert_env_vars_or_raise_error(self.env_vars) if self.env_vars else []
if self.pod_runtime_info_envs:
self.env_vars.extend(self.pod_runtime_info_envs)
if self.pod_template_file:
self.log.debug("Pod template file found, will parse for base pod")
pod_template = pod_generator.PodGenerator.deserialize_model_file(self.pod_template_file)
if self.full_pod_spec:
pod_template = PodGenerator.reconcile_pods(pod_template, self.full_pod_spec)
elif self.pod_template_dict:
self.log.debug("Pod template dict found, will parse for base pod")
pod_template = pod_generator.PodGenerator.deserialize_model_dict(self.pod_template_dict)
if self.full_pod_spec:
pod_template = PodGenerator.reconcile_pods(pod_template, self.full_pod_spec)
elif self.full_pod_spec:
pod_template = self.full_pod_spec
else:
pod_template = k8s.V1Pod(metadata=k8s.V1ObjectMeta())
pod = k8s.V1Pod(
api_version="v1",
kind="Pod",
metadata=k8s.V1ObjectMeta(
namespace=self.namespace,
labels=self.labels,
name=self.name,
annotations=self.annotations,
),
spec=k8s.V1PodSpec(
node_selector=self.node_selector,
affinity=self.affinity,
tolerations=self.tolerations,
init_containers=self.init_containers,
host_aliases=self.host_aliases,
containers=[
k8s.V1Container(
image=self.image,
name=self.base_container_name,
command=self.cmds,
ports=self.ports,
image_pull_policy=self.image_pull_policy,
resources=self.container_resources,
volume_mounts=self.volume_mounts,
args=self.arguments,
env=self.env_vars,
env_from=self.env_from,
security_context=self.container_security_context,
termination_message_policy=self.termination_message_policy,
)
],
image_pull_secrets=self.image_pull_secrets,
service_account_name=self.service_account_name,
host_network=self.hostnetwork,
hostname=self.hostname,
subdomain=self.subdomain,
security_context=self.security_context,
dns_policy=self.dnspolicy,
dns_config=self.dns_config,
scheduler_name=self.schedulername,
restart_policy="Never",
priority_class_name=self.priority_class_name,
volumes=self.volumes,
active_deadline_seconds=self.active_deadline_seconds,
),
)
pod = PodGenerator.reconcile_pods(pod_template, pod)
if not pod.metadata.name:
pod.metadata.name = create_unique_id(
task_id=self.task_id, unique=self.random_name_suffix, max_length=POD_NAME_MAX_LENGTH
)
elif self.random_name_suffix:
# user has supplied pod name, we're just adding suffix
pod.metadata.name = add_unique_suffix(name=pod.metadata.name)
if not pod.metadata.namespace:
hook_namespace = self.hook.get_namespace()
pod_namespace = self.namespace or hook_namespace or self._incluster_namespace or "default"
pod.metadata.namespace = pod_namespace
for secret in self.secrets:
self.log.debug("Adding secret to task %s", self.task_id)
pod = secret.attach_to_pod(pod)
if self.do_xcom_push:
self.log.debug("Adding xcom sidecar to task %s", self.task_id)
pod = xcom_sidecar.add_xcom_sidecar(
pod,
sidecar_container_image=self.hook.get_xcom_sidecar_container_image(),
sidecar_container_resources=self.hook.get_xcom_sidecar_container_resources(),
)
labels = self._get_ti_pod_labels(context)
self.log.info("Building pod %s with labels: %s", pod.metadata.name, labels)
# Merge Pod Identifying labels with labels passed to operator
pod.metadata.labels.update(labels)
# Add Airflow Version to the label
# And a label to identify that pod is launched by KubernetesPodOperator
pod.metadata.labels.update(
{
"airflow_version": airflow_version.replace("+", "-"),
"airflow_kpo_in_cluster": str(self.hook.is_in_cluster),
}
)
pod_mutation_hook(pod)
return pod
[docs] def dry_run(self) -> None:
"""
Print out the pod definition that would be created by this operator.
Does not include labels specific to the task instance (since there isn't
one in a dry_run) and excludes all empty elements.
"""
pod = self.build_pod_request_obj()
print(yaml.dump(prune_dict(pod.to_dict(), mode="strict")))
[docs] def process_duplicate_label_pods(self, pod_list: list[k8s.V1Pod]) -> k8s.V1Pod:
"""
Patch or delete the existing pod with duplicate labels.
This is to handle an edge case that can happen only if reattach_on_restart
flag is False, and the previous run attempt has failed because the task
process has been killed externally by the cluster or another process.
If the task process is killed externally, it breaks the code execution and
immediately exists the task. As a result the pod created in the previous attempt
will not be properly deleted or patched by cleanup() method.
Return the newly created pod to be used for the next run attempt.
"""
new_pod = pod_list.pop(self._get_most_recent_pod_index(pod_list))
old_pod = pod_list[0]
self.patch_already_checked(old_pod, reraise=False)
if self.on_finish_action == OnFinishAction.DELETE_POD:
self.process_pod_deletion(old_pod)
return new_pod
@staticmethod
def _get_most_recent_pod_index(pod_list: list[k8s.V1Pod]) -> int:
"""Loop through a list of V1Pod objects and get the index of the most recent one."""
pod_start_times: list[datetime.datetime] = [
pod.to_dict().get("status").get("start_time") for pod in pod_list
]
most_recent_start_time = max(pod_start_times)
return pod_start_times.index(most_recent_start_time)
class _optionally_suppress(AbstractContextManager):
"""
Returns context manager that will swallow and log exceptions.
By default swallows descendents of Exception, but you can provide other classes through
the vararg ``exceptions``.
Suppression behavior can be disabled with reraise=True.
:meta private:
"""
def __init__(self, *exceptions, reraise: bool = False) -> None:
self._exceptions = exceptions or (Exception,)
self.reraise = reraise
self.exception = None
def __enter__(self):
return self
def __exit__(self, exctype, excinst, exctb) -> bool:
error = exctype is not None
matching_error = error and issubclass(exctype, self._exceptions)
if (error and not matching_error) or (matching_error and self.reraise):
return False
elif matching_error:
self.exception = excinst
logger = logging.getLogger(__name__)
logger.exception(excinst)
return True