Make the helm chart more standard

[chevdor]

While reviewing #368, I noticed a few point that I gathered here.

• indeed, docker-compose seems to be working fine
• While testing the helm chart, I noticed that the current version has quite some strong requirements on an existing infra (See affinity for instance). I see that many sections have been commented out, IMO affinity can be present but should also be commented out by default
• Most defaults in values.yaml are really parity specifc
• I don't have a LoadBalancer in my cluster. It is quite easy to pass values so the shard service uses NodePort. Setting this as default, may allow more users to have it work right away.
• after fixing the above, almost everything started fine expect an issue with the readiness probe of the frontend: Readiness probe failed: Get "http://10.244.0.184:8000/": dial tcp 10.244.0.184:8000: connect: connection refused and the frontend is stuck in a crash loop. This issue is due to the fact that I switched to a NodePort but the template only considers targetPort so the 2 probes are testing on the wrong port. A quick fix is to adjust the targetPort that makes it a bit hacky
• IMO value.yml should work for a "neutral" cluster such as a brand new minikube. We could have a value-parity.yml in the repo for our infra but that may not be the ideal location for such a file
• In the same line I think the frontend could be exposed as NodePort by default

After fixing the above, I could install the chart. However, my http requests to frontend and shard failed from outside the cluster. A curl from inside the cluster from a test pod did repond though.

So here is my test setup:

• comment out the affinity section
• change the service values as shown below

service:
  shard:
    type: NodePort
    port: 80
    targetPort: 80
    nodePort: 30666
    annotations:
      external-dns.alpha.kubernetes.io/hostname: submit.telemetry.parity-stg.parity.io.
  core:
    type: ClusterIP
    port: 80
    targetPort: 8000
    annotations: {}
    #  cloud.google.com/load-balancer-type: Internal
    #  networking.gke.io/internal-load-balancer-allow-global-access: "true"
    #  external-dns.alpha.kubernetes.io/hostname: feed.telemetry.parity-stg.parity.io.
  frontend:
    type: NodePort
    port: 80
    targetPort: 80
    nodePort: 30667
    annotations: {}

[chevdor]

See also #369

[chevdor]

I am closing this issue as there is now a new location where the issue will be better tracked.