[
https://issues.apache.org/jira/browse/AIRFLOW-3214?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Zhi Lin updated AIRFLOW-3214:
-----------------------------
Description:
I have a fresh install of airflow on Kubernetes, basically running scripts in
scripts/ci/kubernetes, and I fix the flask-appbuilder version so the webserver
is running fine.
But when I try to run the example-kubernetes-executor dag, it fails and shows
nothing useful in the log. Then I try to pull the image airflow/ci manually,
and also fails. It says no such image or login is needed. I try to change the
image in the dag to something I already have locally, but it still fails and
this time no log...
+++
Updated. The reason is that airflow is needed in that image specified. I set
deleteworkerpod to false to discover this.
+++
And I wonder if I want to use KubernetesPodOperator, I cannot run airflow in a
k8s pod? Because it tries to find the config files of k8s...
So what should I do? What I want to do is to run some machine learning task in
a image specified. Sometimes I need to run a python script in that spawned pod,
so I guess I need to put that file in the dag folder? Any suggestion
appreciated.
was:
I have a fresh install of airflow on Kubernetes, basically running scripts in
scripts/ci/kubernetes, and I fix the flask-appbuilder version so the webserver
is running fine.
But when I try to run the example-kubernetes-executor dag, it fails and shows
nothing useful in the log. Then I try to pull the image airflow/ci manually,
and also fails. It says no such image or login is needed. I try to change the
image in the dag to something I already have locally, but it still fails and
this time no log...
And I wonder if I want to use KubernetesPodOperator, I cannot run airflow in a
k8s pod? Because it tries to find the config files of k8s...
So what should I do? What I want to do is to run some machine learning task in
a image specified. Sometimes I need to run a python script in that spawned pod,
so I guess I need to put that file in the dag folder? Any suggestion
appreciated.
> [Kubernetes] No logs, hard to debug
> -----------------------------------
>
> Key: AIRFLOW-3214
> URL: https://issues.apache.org/jira/browse/AIRFLOW-3214
> Project: Apache Airflow
> Issue Type: Bug
> Components: kubernetes
> Affects Versions: 1.10.0
> Reporter: Zhi Lin
> Priority: Trivial
>
> I have a fresh install of airflow on Kubernetes, basically running scripts in
> scripts/ci/kubernetes, and I fix the flask-appbuilder version so the
> webserver is running fine.
> But when I try to run the example-kubernetes-executor dag, it fails and shows
> nothing useful in the log. Then I try to pull the image airflow/ci manually,
> and also fails. It says no such image or login is needed. I try to change the
> image in the dag to something I already have locally, but it still fails and
> this time no log...
> +++
> Updated. The reason is that airflow is needed in that image specified. I set
> deleteworkerpod to false to discover this.
> +++
> And I wonder if I want to use KubernetesPodOperator, I cannot run airflow in
> a k8s pod? Because it tries to find the config files of k8s...
> So what should I do? What I want to do is to run some machine learning task
> in a image specified. Sometimes I need to run a python script in that spawned
> pod, so I guess I need to put that file in the dag folder? Any suggestion
> appreciated.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
