Kubernetes Helm chart

[nodesocket]

I see there is support for Docker, any chance there is a Kubernetes Helm chart or one can be created?

[mitchdenny]

I haven't used a Helm chart (I sometimes find them to be overkill), but here is the deployment file that I use in my little home lab:

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: zero-git-pvc
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 10Gi
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: zero-git
  labels:
    app: zero-git 
spec:
  selector:
    matchLabels:
      app: zero-git
  serviceName: zero-git
  template:
    metadata:
      labels:
        app: zero-git
    spec:
      containers:
        - name: zero-git
          image: charmcli/soft-serve:latest
          volumeMounts:
            - name: zero-git-volume
              mountPath: /soft-serve 
      volumes:
        - name: zero-git-volume
          persistentVolumeClaim:
            claimName: zero-git-pvc
---
apiVersion: v1
kind: Service
metadata:
  name: zero-git
spec:
  selector:
    app: zero-git
  type: LoadBalancer
  externalTrafficPolicy: Cluster
  ports:
    - name: ssh
      protocol: TCP
      port: 22
      targetPort: 23231

Note that I'm exposing a load balanced service on port 22 in my environment because I'm making use of MetalLB to provide layer 7 load balancing which is why I can use port 22 without conflicting with ports in use on the K8s node. I'm also using https://longhorn.io to provide the volume to the pod in the statefulset so that if I take down one of the nodes in the cluster the volume will be transparently available to the underlying pod when it launches on another node.

[rigzba21]

I was running into issues in my deployment if I didn't set the environment variables SOFT_SERVE_PORT and SOFT_SERVE_BIND_ADDRESS. The container was giving an error:

2022/03/12 16:02:06 env: parse error on field "Port" of type "int": strconv.ParseInt: parsing "tcp://10.43.76.7:22": invalid syntax

until I set those two environment variables.

Here's my setup:

• k3s with default local-path StorageClass

• MetalLB

  apiVersion: apps/v1
  kind: StatefulSet
  metadata:
  name: soft-serve-deployment
  namespace: soft-serve
  labels:
  app: soft-serve
  spec:
  selector:
  matchLabels:
    app: soft-serve
  serviceName: soft-serve      
  template:
  metadata:
    labels:
      app: soft-serve
  spec:
    volumes:
    - name: git-repo-data
      persistentVolumeClaim:
        claimName: soft-serve-pvc
    containers:
      - name: soft-serve
        image: docker.io/charmcli/soft-serve:latest
        ports:
        - containerPort: 23231
        env:
        - name: SOFT_SERVE_PORT
          value: '23231'
        - name: SOFT_SERVE_BIND_ADDRESS
          value: '0.0.0.0'
        volumeMounts:
        - name: git-repo-data
          mountPath: /soft-serve              
  ---
  apiVersion: v1
  kind: PersistentVolumeClaim
  metadata:
  name: soft-serve-pvc
  namespace: soft-serve
  spec:
  resources:
      requests:
          storage: 5Gi
  accessModes:
      - ReadWriteOnce
  storageClassName: local-path
  ---
  apiVersion: v1
  kind: Service
  metadata:
  name: soft-serve
  namespace: soft-serve
  spec:
  type: LoadBalancer
  externalTrafficPolicy: Cluster
  selector:
  app: soft-serve
  ports:
  
  - name: ssh
    protocol: TCP
    port: 22 
    targetPort: 23231

  Hope this helps!

[johndietz]

agree that a helm install would be clutch. appreciate the quick start rigzba21 - for anyone that ran that and it didn't work double-check your pvc service class and adjust local-path to an available one shown from kubectl get sc

[roertel]

I see there is support for Docker, any chance there is a Kubernetes Helm chart or one can be created?

I started an initial Helm chart, but haven't had time to thoroughly test it. I'm still working out how to expose the port. Feel free to MR/PR/Fork or send bug reports or ideas.

[brizzbuzz]

has anyone gotten this working on a remote cluster? can run soft serve in docker locally no problem, but when I try to deploy to a cluster (no pvc yet, just deployment + svc) I get kex_exchange_identification: read: Connection reset by peer

[brizzbuzz]

just a follow up here, my pod seems to be crash looping for some reason (asserted this by adding the promwish middleware and using as a liveness check). this in turn was causing my known hosts to be incorrect, though why it didn't show the normal known hosts warning and instead caused a kex_exchange error I am not sure