Context
When developing your application or process on EDITO, you might run into some unexpected behavior for your chart.
💡 EDITO Pro Tip: from the EDITO Datalab, you can easily access:
Your namespace events by clicking on the “Events” button
(1)from “My Applications” or “My processes”A running application or process logs by clicking on the name
(2)on its tile
⚠️ Logs disappear with the pod. Once a process has ended, successfully or not, its pod is removed with its logs, and only the events remain. If you need to investigate a failure later, read the logs while the process is still running, or make your code write its errors to your storage.
⚠️ Failed processes are not cleaned up automatically. A failed process stays in “My processes” with the status `failed` until you delete it, and keeps counting in your quotas.
💡EDITO Pro Tip: you can also follow the events from a script, see Follow your applications and processes.
Check from inside a service
Still, to understand better what is going on under the hood, you can also run some commands from inside a service that could help you. We will show some basic examples, but you can always learn more in the official Kubernetes documentation.
To do so, you need a terminal on EDITO with the “edit” or “admin” Kubernetes roles: for example, you can launch a Jupyter-python with “edit” role.
First, you can start by listing the pods running in your namespace:
kubectl get pods
Then, you can print a detailed description of a selected pod by running:
kubectl describe pods/my-pod-xxxxxx-x
It can be useful if your pod does not start.
If your pod has started, you can get its logs by running:
kubectl logs my-pod-xxxxxx-x --all-containers=true
📌 Note: you can use the --container container-name option to target a specific container.
Sometimes logs aren't enough, and you would want to run commands directly on your container. This is possible with the kubectl exec command. To get a shell in your running container, simply run:
kubectl exec --stdin --tty my-pod -- /bin/bash
📌 Note: this assume bash is available in your container, but any other available shell would work.
If your application/process won’t launch from the Datalab
You might run into an issue where your application/process won’t launch from the Datalab, without an explicit error to debug.
If you cannot even access the launcher page of your application/process, you probably have a malformed chart. You can try to open the debug console of your browser which might hint you where the error might be.
If you can access the launcher page but cannot launch your application/process, you might try to manually install your chart.
First, you need a terminal on EDITO with the “edit” or “admin” Kubernetes roles; for example, you can launch a Jupyter-python with “edit” role.
Then, you can go on the launcher page of the service/process you are developing and expand the command bar showing Helm commands. There should be three commands:
The first one to add the Helm repository, starting with
helm repo add ...The second one creating a file with your values, starting with
cat <<EOF > ./values.yaml ...The last one installing your chart, starting with
helm install ...
Copy and run those three commands in the Jupyter-python terminal, and the last one should give you a better hint as to why your application/process is crashing.
If necessary (for example in the case of ghost processes that use your computing resources), you can run the command
helm uninstallto kill your ghost processes or applications.
What's next?
If you have any questions, problems, or suggestions, please feel free to contact us via chat using the widget available at the bottom right of the page.

