Zuchens opened a new issue #9966:
URL: https://github.com/apache/airflow/issues/9966
**Apache Airflow version**: 1.10.11
**Kubernetes version (if you are using kubernetes)** (use `kubectl
version`): 1.16
**Environment**:
- **Cloud provider or hardware configuration**: AKS
- **OS** (e.g. from /etc/os-release):
- **Kernel** (e.g. `uname -a`):
- **Install tools**: helm chart
- **Others**: KubernetesExecutor
**What happened**:
We have set up environment variable:
AIRFLOW__KUBERNETES_LABELS__TYPE='airflow-task-pod' that should create label in
airflow configuration.
After ugrading airflow from 1.10.10 to 1.10.11, custom kubernetes_label in
config stopped appearing in the logs.
Previous sample from kibana:
```
"kubernetes": {
"labels": {
"try_number": "1",
"airflow-worker": "xxx",
"airflow_version": "1.10.10",
"dag_id": "xxx",
"kubernetes_executor": "True",
"task_id": "xxx",,
"execution_date": "2020-07-21T02_10_00_plus_00_00",
"type": "airflow-task-pod"
},
}
```
Current sample:
```
"kubernetes": {
"labels": {
"try_number": "1",
"airflow_version": "1.10.11",
"task_id": "xxx",
"dag_id": "xxx",
"kubernetes_pod_operator": "True",
"execution_date": "2020-07-21T0210000000-d4ef2bace"
}
}
```
**What you expected to happen**:
After upgrading logs should stay the same as in airflow 1.10.10.
**What do you think went wrong?**:
The label appears in
airflow.configuration.as_dict().get('kubernetes_labels') the issue seems to be
in PodGenerator.
PodGenerator.construct_pod() will ignore any labels from
worker_configuration since they will always be replaced by dynamic_pod
configuration labels.
**How to reproduce it**:
Here is a locally run snipet to show the behaviour:
```
import json
from airflow.kubernetes.pod_generator import PodGenerator
from kubernetes.client import V1ObjectMeta, V1Pod
pod = PodGenerator.construct_pod(
namespace="test",
worker_uuid="test",
pod_id="test",
dag_id="test",
task_id="test",
try_number=1,
date="23-07-2020",
command="test",
kube_executor_config=None,
worker_config=V1Pod(metadata=V1ObjectMeta(labels={"type":
"airflow-task-pod"})))
print(json.dumps(pod.metadata.labels, indent=2))
```
If this is expected behaviour can you suggest me other way to handle custom
labels? :)
Thanks!
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
