Harden elasticsearch chart for Kube 1.5

[icereval]

Further extends upon PR #890

• Added configmap to explicitly provide cluster configurations and scripts

• Replace depreciating ES_HEAP_SIZE with ES_JAVA_OPTS to position for ES v5 support

• Removed alpha storage class operators

• Removed catastrophic liveness probe checking entire clusters health

• Readiness probe now inspects local node health

• Added termination grace period (defaults to 60m) to allow pre-stop-script.sh time to gracefully migrate shards

• Added init container to configure vm.max_map_count

• Updated elasticsearch.yaml:

  • Added PROCESSOR configuration to prevent large cluster garbage collection issues leading to node eviction
  • Added configurable gateway defaults to help avoid a split brain, requiring two masters online and in consensus before recovery can continue

• Updated pre-stop-script.sh:

  • Check v1beta1 statefulset endpoint
  • Evalute .spec.replicas for statefulset desired size
  • Clear _cluster/settings ip exclusion prior to shutdown to avoid a possible (random) ip match scenario on expansion of the clsuter

• Data nodes now use default storage class if one is not specified

• Apply Helm best practices to prep for stable

[k8s-ci-robot]

Hi @icereval. Thanks for your PR.

I'm waiting for a kubernetes member to verify that this patch is reasonable to test. If it is, they should reply with @k8s-bot ok to test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here.

[unguiculus]

I would prefer to not move the chart to stable as long as config values start with an uppercase letter. In this case, I would also switch to nesting first. I think this make sense here, e. g.:

client:
  replicas: 2
  port: 2379
  heapSize: "128m"
master:
  replicas: 2
  heapSize: "128m"
data:
  replicas: 3
  storage: "30Gi"

See https://github.com/kubernetes/helm/blob/master/docs/chart_best_practices/values.md

Generally, I think a move to stable should be a separate PR so it is explicit.

Regarding resources, it's becoming increasingly popular to specify resources like so (without specifying defaults): https://github.com/kubernetes/charts/blob/master/stable/nginx-ingress/templates/controller-deployment.yaml#L93

[icereval]

I completely agree and all are very good points. Would you mind informing the base PR #890, if we can get this merged, even if still in incubator status one of us can follow up with a new PR and the final changes needed to make this stable.

[viglesiasce]

@simonswine @unguiculus any update on this review? Were the latest changes enough to merge?

[unguiculus]

As suggested in #890, a move to stable should only happen after applying best practices, e. g. regarding values. See https://github.com/kubernetes/helm/blob/master/docs/chart_best_practices/values.md.

[icereval]

@unguiculus & @prydonius, rebase complete and a first pass at helm best practices applied to the incubator chart.

[unguiculus]

Excellent. This goes in the right direction. Now, please have a look at app labels, which should be {{ template "name" . }}. Note that you will then have to add the release label to the selector in services. nginx-ingress an excellent example:

https://github.com/kubernetes/charts/blob/master/stable/nginx-ingress/templates/controller-deployment.yaml#L6
https://github.com/kubernetes/charts/blob/master/stable/nginx-ingress/templates/controller-service.yaml#L56-L58

[unguiculus]

@icereval Are you working on an update?

[icereval]

@unguiculus, yep, I'll have my updates ready soon

[icereval]

@unguiculus, best practice changes pushed up

[unguiculus]

Would you mind adding a NOTES.txt? Otherwise it looks nice but I have yet to review more thoroughly. I installed it on GKE and everything came up nicely.

[unguiculus]

@k8s-bot ok to test

[icereval]

Added NOTES.txt and client.serviceType based on review of the concourse stable chart.