Potential update for helm test

[lachie83]

The idea behind helm test is to provide a simple red/green signal a successful chart install. In the case of consul - the health of the cluster. The way helm tests are implemented are via run to complete or (Jobs) with exit codes. We currently have the following (see below) which looks similar to the upstream chart. This test is broken due to rbac (we can fix that) but I'm wondering if there is a better way to run the test without the need to access the Kubernetes API. Potentially like the readiness probe? Looking for suggestions.

$ helm test $(helm last)
RUNNING: virulent-chicken-test-mtu5z
FAILED: virulent-chicken-test-mtu5z, run `kubectl logs virulent-chicken-test-mtu5z --namespace default` for more info
Error: 1 test(s) failed

laevenso@rue [(⎈ |1-10:default)] ~/sandbox/consul-k8s/helm on master
$ kubectl logs virulent-chicken-test-mtu5z --namespace default
1..1
not ok 1 Testing Consul cluster has quorum
# (in test file /tests/run.sh, line 2)
#   `[ `kubectl exec virulent-chicken-consul-server-0 consul members --namespace=default | grep server | wc -l` -ge "3" ]' failed
# Error from server: pods "virulent-chicken-consul-server-0" is forbidden: User "system:serviceaccount:default:default" cannot get pods in the namespace "default"

[mitchellh]

Yeah, this has been on my TODO as well. There are a couple tests we can do:

1.) We can do a simple agent HTTP call, if you have client nodes enabled. This doesn't require any perms, just has to be deployed to the correct place.

2.) If client nodes aren't installed, we can use service discovery to the servers.

We can probably do both. Noted!

[mitchellh]

Fixed this by using the Consul agent API. Its somewhat brittle in that it only works if you have Consul agents everywhere, but we can fix that up over time.