Add custom values file for scalardb-graphql Helm chart

[ymorimo]

This PR adds a custom values file for scalardb-graphql Helm chart.

This file is going to be referenced from the deployment guide. scalar-labs/scalardb-graphql#18

Currently, it covers AWS only. The database configurations and the Ingress annotations for other services should be added when needed.

[kota2and3kan]

Overall looks good to me, but I have comments about tolerations and podAntiAffinity.

---

1. About tolerations

It seems that there is no configuration of tolerations like scalardl-custom-values.yaml and scalardb-custom-values.yaml.

https://github.com/scalar-labs/scalar-kubernetes/blob/master/conf/scalardl-custom-values.yaml#L99-L103
https://github.com/scalar-labs/scalar-kubernetes/blob/master/conf/scalardb-custom-values.yaml#L86-L90

I think some non Scalar DB GraphQL Server pods (e.g. user application pod) will be deployed to the node group for Scalar DB GraphQL Server, if there is no configuration of tolerations.
So, I think it is better to add tolerations to this sample custom values file.
(Also, in this case, we need to add the section that explain about adding Taints to the Kubernetes worker node in the deployment guide.)
What do you think?

---

2. About podAntiAffinity

It seems that there is no configuration of podAntiAffinity like scalardl-custom-values.yaml and scalardb-custom-values.yaml.
https://github.com/scalar-labs/scalar-kubernetes/blob/master/conf/scalardb-custom-values.yaml#L101-L115
https://github.com/scalar-labs/scalar-kubernetes/blob/master/conf/scalardl-custom-values.yaml#L114-L128

I think several Scalar DB GraphQL pods will be deployed to the same one node, if there is no configuration of podAntiAffinity.
So, I think it is better to add podAntiAffinity to distribute pods within the Kubernetes cluster.
What do you think?

[ymorimo]

@kota2and3kan Thank you for the suggestion. I didn't fully understand how tolerations and podAntiAffinity work, but now they are clearer thanks to your explanation.
I have added them in ff82496 and I'm going to add a description about Taints to the deployment guide right away.

[kota2and3kan]

@ymorimo

I have added them in ff82496 and I'm going to add a description about Taints to the deployment guide right away.

Thank you for the update!
For your reference, I created the PR about Taints in #183.

[ymorimo]

@kota2and3kan Sorry for the late reply. I have added a description about Taints in scalar-labs/scalardb-graphql#32. Please take a look.

[kota2and3kan]

LGTM! Thank you!

---

It worked well in my local environment!

• tolerations

  $ kubectl describe node minikube | grep Taints
  Taints:             <none>
  
  $ kubectl describe node minikube-m02 | grep Taints
  Taints:             kubernetes.io/app=scalardbgraphqlpool:NoSchedule

  $ kubectl get pod -o wide
  NAME                                READY   STATUS    RESTARTS   AGE   IP           NODE           NOMINATED NODE   READINESS GATES
  mysql-scalardb-0                    1/1     Running   0          50m   172.17.0.4   minikube       <none>           <none>
  postgresql-scalardb-0               1/1     Running   0          50m   172.17.0.5   minikube       <none>           <none>
  scalardb-6644875756-28fld           1/1     Running   0          48m   172.17.0.7   minikube       <none>           <none>
  scalardb-envoy-84bf44bc5d-84c88     1/1     Running   0          48m   172.17.0.6   minikube       <none>           <none>
  scalardb-graphql-7db6cdd8b5-8596x   0/1     Running   0          2s    172.17.0.4   minikube-m02   <none>           <none>
  scalardb-graphql-7db6cdd8b5-vg9k5   0/1     Running   0          3s    172.17.0.3   minikube-m02   <none>           <none>
  scalardb-graphql-7db6cdd8b5-zls4b   0/1     Running   0          2s    172.17.0.2   minikube-m02   <none>           <none>

  • Scalar DB GraphQL pods run on minikube-m02 and other pods run on minikube. In other words, I was able to make minikube-m02 dedicated node for Scalar DB GraphQL.

• podAntiAffinity (I use requiredDuringSchedulingIgnoredDuringExecution for testing.)

  $ kubectl get pod -o wide
  NAME                                READY   STATUS    RESTARTS   AGE   IP           NODE           NOMINATED NODE   READINESS GATES
  mysql-scalardb-0                    1/1     Running   0          52m   172.17.0.4   minikube       <none>           <none>
  postgresql-scalardb-0               1/1     Running   0          51m   172.17.0.5   minikube       <none>           <none>
  scalardb-6644875756-28fld           1/1     Running   0          49m   172.17.0.7   minikube       <none>           <none>
  scalardb-envoy-84bf44bc5d-84c88     1/1     Running   0          49m   172.17.0.6   minikube       <none>           <none>
  scalardb-graphql-7bf5c4744c-52gc4   0/1     Pending   0          2s    <none>       <none>         <none>           <none>
  scalardb-graphql-7bf5c4744c-qrcpv   0/1     Running   0          2s    172.17.0.2   minikube-m02   <none>           <none>
  scalardb-graphql-7bf5c4744c-xfmtt   0/1     Pending   0          2s    <none>       <none>         <none>           <none>

  • Only one Scalar DB GraphQL pod runs on one node.

[feeblefakie]

Left a minor suggestion.
Other than that, LGTM! Thank you!

[brfrn169]

LGTM! Thank you!