Helm is not pre-installed in AKS clusters

[rspaulino]

https://docs.microsoft.com/en-us/azure/aks/kubernetes-helm

Base on the documentation it state that Helm is pre-installed in AKS clusters but when I run

helm install stable/nginx-ingress
Error: could not find tiller

I get an error and when I look at the subsystem installed in k8s Tiller isn't listed.

kubectl get pods --namespace kube-system
NAME                                    READY     STATUS             RESTARTS   AGE
heapster-b5ff6c4dd-ztxsb                2/2       Running            0          18m
kube-dns-v20-6c8f7f988b-hsq25           3/3       Running            0          18m
kube-dns-v20-6c8f7f988b-knczn           3/3       Running            0          18m
kube-proxy-9jc4t                        1/1       Running            0          15m
kube-proxy-tx6pd                        1/1       Running            0          15m
kube-svc-redirect-96tt7                 0/1       CrashLoopBackOff   7          15m
kube-svc-redirect-9nw5l                 0/1       CrashLoopBackOff   7          15m
kubernetes-dashboard-7f7d9489fc-rzlxt   0/1       CrashLoopBackOff   6          18m
tunnelfront-6mwfv                       0/1       CrashLoopBackOff   7          15m
tunnelfront-glvmz                       0/1       CrashLoopBackOff   7          15m

[seanknox]

Good call out that should be quick for us to ship soon.

[ghost]

I was kinda surprised to see helm isn't installed by default, and I'm having trouble setting it up myself

[rspaulino]

@k1-hedayati what problem you having? once I got the credentials I ran helm command and tiller was install on the cluster.

[ghost]

@rspaulino my issue was because of #2, I've created a new cluster on uk and I could install helm manually, thanks

[rspaulino]

@k1-hedayati Yes, I must say I'm not happy with this release it's been buggy but my biggest concern is that this is the cloud, it shouldn't be not capacity issues since allocating servers should be almost automatically but I know it's beta.

[dtzar]

We just provisioned two new clusters and both of them we were able to install helm init with v2.7.0 to AKS but it seems helm is unable to connect to the cluster. helm version returns:
Error: cannot connect to Tiller even though tiller is running:

kube-svc-redirect-9rr9f                 0/1       CrashLoopBackOff   7          16m
kube-svc-redirect-cr48w                 0/1       CrashLoopBackOff   6          16m
kube-svc-redirect-wcmz7                 0/1       CrashLoopBackOff   7          16m
kubernetes-dashboard-6fc8cf9586-jdgwz   0/1       CrashLoopBackOff   7          16m
tiller-deploy-84b97f465c-sjj29          1/1       Running            0          50s
tunnelfront-2qfqc                       0/1       CrashLoopBackOff   7          16m
tunnelfront-jhdw8                       0/1       CrashLoopBackOff   7          16m
tunnelfront-ntrm4                       0/1       CrashLoopBackOff   7          16m

tiller logs state:

Error from server: Get https://aks-agentpool-55225649-1:10250/containerLogs/kube-system/tiller-deploy-84b97f465c-sjj29/tiller: dial tcp 10.240.
0.6:10250: getsockopt: connection refused

[RehanSaeed]

Helm 3 does away with Tiller, so this is no longer needed. I think this issue can be closed.

[kvolkovich-sc]

Completely agree.

Cluster-wide Tiller in kube-system namespace is a big hole for security.

But you don't need a Tiller for Helm 2 too.

Just use something similar to this script:

#!/usr/bin/env bash

# Runs Tiller locally, then executes any Helm command
# E.g. no need to install Tiller server within Kubernetes cluster
# Examples:
# helm-serverless ls
# helm-serverless version

tiller -storage=secret &> /tmp/tiller.log & tiller_pid="$!"

function finish {
	kill "$tiller_pid"
}
trap finish EXIT


export HELM_HOST=":44134"
helm $@
helm_exit_code=$?

exit $helm_exit_code

[jnoller]

So, helm & tiller are rolled into AKS - when Helm 3 is released, we will examine removing the server side tiller bits.