Which chart:
Discourse 4.2.3
Describe the bug
Long story short I found that the helm chart adds a port to the DISCOURSE_HOST in the configmap if the port is not 80. In this case I am using 443 for SSL offloading on a AWS NLB. This breaks avatar images from working after you upload them. If I remove the port from the configmap directly on the cluster that fixes the issue. But any images uploaded before the change can't be used again. Even spent time trying to delete them from the file system and clearing assets and running precompile again. Spent way to much time on this. The problem is unless you change the SHA of an image you can never upload that image again. Even if I reverted the change but then images don't load.
|
DISCOURSE_HOST: "{{ include "discourse.host" . }}{{- if ne $port "80" }}:{{ .Values.service.port }}{{ end }}" |
Line 14 is where it adds the port in. From what I can gather there shouldn't be a port there at all, this value is just the hostname of the instance of discourse.
I think removing that port should work. I can do a PR tomorrow if that won't break anything else?
To Reproduce
Steps to reproduce the behavior:
Use port 443
- deploy discourse
- upload a logo, avatar, etc.
- Edit the configmap to remove the extra port in DISCOURSE_HOST
- Restart pod
- Previous images don't show so clear and reupload and they still won't work
- Change the image or upload a different image and those work fine
Expected behavior
Image/Avatars work out of the box, including profile menu icons etc.
Version of Helm and Kubernetes:
Kubernetes 1.19 on EKS
version.BuildInfo{Version:"v3.1.1", GitCommit:"afe70585407b420d0097d07b21c47dc511525ac8", GitTreeState:"clean", GoVersion:"go1.13.8"}
- Output of
kubectl version:
Client Version: version.Info{Major:"1", Minor:"18", GitVersion:"v1.18.2", GitCommit:"52c56ce7a8272c798dbc29846288d7cd9fbae032", GitTreeState:"clean", BuildDate:"2020-04-16T23:34:25Z", GoVersion:"go1.14.2", Compiler:"gc", Platform:"darwin/amd64"}
Server Version: version.Info{Major:"1", Minor:"19+", GitVersion:"v1.19.8-eks-96780e", GitCommit:"96780e1b30acbf0a52c38b6030d7853e575bcdf3", GitTreeState:"clean", BuildDate:"2021-03-10T21:32:29Z", GoVersion:"go1.15.8", Compiler:"gc", Platform:"linux/amd64"}
Additional context
Add any other context about the problem here.
Using AWS Network load balancer with ACM on port 443
Which chart:
Discourse 4.2.3
Describe the bug
Long story short I found that the helm chart adds a port to the
DISCOURSE_HOSTin the configmap if the port is not80. In this case I am using 443 for SSL offloading on a AWS NLB. This breaks avatar images from working after you upload them. If I remove the port from the configmap directly on the cluster that fixes the issue. But any images uploaded before the change can't be used again. Even spent time trying to delete them from the file system and clearing assets and running precompile again. Spent way to much time on this. The problem is unless you change the SHA of an image you can never upload that image again. Even if I reverted the change but then images don't load.charts/bitnami/discourse/templates/configmaps.yaml
Line 14 in 131fa45
Line 14 is where it adds the port in. From what I can gather there shouldn't be a port there at all, this value is just the hostname of the instance of discourse.
I think removing that port should work. I can do a PR tomorrow if that won't break anything else?
To Reproduce
Steps to reproduce the behavior:
Use port 443
Expected behavior
Image/Avatars work out of the box, including profile menu icons etc.
Version of Helm and Kubernetes:
Kubernetes 1.19 on EKS
helm version:kubectl version:Additional context
Add any other context about the problem here.
Using AWS Network load balancer with ACM on port 443