Helm chart stuck in starting state

Questions and Answers
1

Hello,

I have a helm chart that is stuck in starting state that I can’t stop :frowning:. Can you have a look please ?

Cluster ID: a5181456-e3ff-46e0-b01d-088c51523656
Organization ID: a2ed680d-ffdf-44a2-905f-24fc2a58a97d
Project ID: 2257e891-1c08-4440-bd76-807e5d1b2d90
Environment ID: 6138a1e6-8090-4c93-9205-004e22473779
Service ID: ee805995-7cb7-4f50-a5fd-069b8ace9220

Thanks a lot :slight_smile:

2

Hello @Orkin ,

I just checked and everything is back to normal.

When a service is in a waiting state, we must wait for a timeout before doing more action.

We can’t cancel it either.

Regards,
Charles-Edouard

3

It was not a timeout problem. The application was deployed properly but the “job” loop failed. I had to stop the cluster for some other reason and it solved the problem :sweat_smile:.

4

Ok,

I just checked but I don’t have any logs on this.

Please let me know if this happens ago and we’ll try to debug it live.

Regards,
Charles-Edouard

1 Like
5

Hi @Orkin,

actually, we might have found the issue.

Your dev cluster is running via Karpenter and there is a known issue with the way it manages Daemonsets (Fluentbit is deployed as a daemonset).

To ensure it is deployed properly on every node of your cluster, you should:

  1. deploy a custom priority class (see this guide)
  2. define the fluentbit chart should use

image
image1388×798 106 KB

1 Like
6

Thanks @a_carrano for your response. Do I need to allow cluster wide ressources for fluent-bit daemonset too ?

7

Hi @Orkin,

the “Allow wider resource” check is run before deploying the resources to ensure the user has the right to deploy resources outside his own namespace (the namespace of the environment where it is deployed).

Fluentbit probably deploys CRDs and pods outside the “environment namespace” so it is necessary to enable the option.