Does not install on AWS about consul-helm

[mitchellh]

Also it is a bit confusing that there exists a chart on helm and one at hashicorp.

Yep, we addressed this in our announcement blog but it isn't documented elsewhere so understandable. We plan on migrating to the official chart in time, but during the first releases we wanted to stabilize here.

As for your actual issue: do you have at least 3 distinct hosts for the servers? The most common case I've seen this pop up is that there aren't enough hosts to place the servers (such as minikube, though I know you stated you're in AWS).

We tested the helm chart on Google, Azure, and AWS hosted K8S offerings and verified that it works with the default settings, so its likely something experimental. We also just spun up another EKS cluster and verified it works out of the box.

If you find out more let me know. Unfortunately there isn't anything else to go on at this point...

[antonyRoberts]

I have 3 machines in my clusters and the setup is working with the chart hosted on helm's site. I'm not using EKS for kubernetes on AWS

Commenting out afinity in statefulset as mentioned in #13 (comment) works for this as well

[mitchellh]

It's possible that for some reason the Consul servers aren't schedulable on all 3 machines (the default replicas is 3 and that'd require 3 machines). Disabling the affinity rules seems to imply that, because the error messages above mean that certain set members didn't appear.

[Vadim-Zenin]

0.2.0 chart version works with EKS.

default       pod/consul-lj8br                     1/1       Running   0          1m
default       pod/consul-pnq7k                     1/1       Running   0          1m
default       pod/consul-server-0                  1/1       Running   0          1m
default       pod/consul-server-1                  1/1       Running   0          1m
default       pod/consul-server-2                  1/1       Running   0          1m
default       pod/consul-twc26                     1/1       Running   0          1m

[mitchellh]

Dope.

We also have a PR to have configurable affinities so that it works with smaller K8S clusters. There is a separate issue tracking that so will close this one.