Add topologySpreadConstraints to the artifactory-ha helm chart

[rkevin-arch]

Is this a request for help?:
No

---

Is this a BUG REPORT or FEATURE REQUEST? (choose one):
Feature Request

It would be nice to support topologySpreadConstraints for the artifactory-ha chart to support spreading Artifactory and nginx pods across multiple nodes / zones. Right now we can technically use podAntiAffinity to force one Artifactory / nginx pod per node, but that's not as flexible as topologySpreadConstraints.

I'm envisioning something along the lines of this in `values.yaml:

artifactory:
  topologySpreadConstraints:
    maxSkew: 1
    whenUnsatisfiable: DoNotSchedule
    topologyKey: kubernetes.io/hostname
nginx:
  topologySpreadConstraints:
    maxSkew: 1
    topologyKey: kubernetes.io/hostname
    whenUnsatisfiable: DoNotSchedule

which gets populated as the following in the primary and member statefulsets:

spec:
  template:
    spec:
      topologySpreadConstraints:
        maxSkew: 1
        topologyKey: kubernetes.io/hostname
        whenUnsatisfiable: DoNotSchedule
        labelSelector:
          matchLabels:
            component: artifactory-ha # or some other label that applies to the primary and all members

and similar for nginx, just with a different labelSelector that matches all nginx pods instead. This guarantees that all Artifactory pods and all nginx pods are spread evenly across all nodes.

Version of Helm and Kubernetes:
Helm 3.8.0, Kubernetes v1.20.11-eks-f17b81

Which chart:
artifactory-ha, although other charts can benefit from this as well

[chukka]

Thanks for this, we will review internally and try to get this added to our backlog and prioritize !

[chukka]

Also, can you please a PR for the request so that it gets to master quickly ?

[dmitrijrub]

im sturggling with 'topologySpreadConstraints:' for 'artifactory' depolyment, is it implemented only for artifactory-ha ?