Upgrade chart to not use ingress controller

[rtisma]

In order for a mutating-webhook-config to work, TLS termination must be enabled on the webhook service. Currently, this is done, via our usual ingress configuration, where certs are automagically created by the cert manager. The issue with this, is that when the kube-apiserver calls the webhook url, it is using the public domain, resulting in hairpinning.

This issue was solved previously, by updating the security groups for the kube-system and kube-apiserver VMs, however since the security groups config involved IPs, when the kube-system kube-apisevers were re-deployed (as part of some maintentance...i think it was the move to SSD), they were assigned new IPs, and thus were not included in the security groups. This resulted in wierd timeout errors when workflow-management was trying to create a wes pod.

One solution proposed by @yalturmes was to create a provision pod which execute the steps outlined in
https://github.com/icgc-argo/rdpc-kube-mutating-webhook/blob/develop/deploy.sh.

Impact

This is critical, since debugging this issue is not trivial and since there are so many moving parts in our env, it is easy to break without this fix

[rosibaj]

• Will be an issue anytime kubernetes maser nodes moves/has ip updated
• Needs to be preventative.