Unable to Uninstall Garden Services From a Cluster

[D-Dinh]

Bug

Current Behavior

When attempting to uninstall garden services from a cluster as documented:
https://docs.garden.io/using-garden/remote-kubernetes

garden --env= plugins kubernetes uninstall-garden-services

The following happens:

Removing cluster-wide services for dev environment ⚙️

✔ providers                 → Getting status... → Done
 Deleting services...
    build-sync                → Deleting...
    default-backend           → Deleting...
    docker-daemon             → Deleting...
    docker-registry           → Deleting...
    ingress-controller        → Deleting...
    nfs-provisioner           → Deleting...
    kubernetes-dashboard      → Deleting...
    registry-proxy            → Deleting...
Error: Command failed with exit code 1 (EPERM)...

Error.log

[2019-10-01T23:40:05.621Z] Error: Command failed with exit code 1 (EPERM): $USER_HOME/.garden/tools/helm/1c6ed8021f49d4ce/darwin-amd64/helm --tiller-namespace garden-system --kube-context kubernetes template --name garden-registry-proxy --namespace garden-system --values /Sandbox/git/my-service/.garden/kubernetes.garden/build/registry-proxy/garden-values.yml /Sandbox/git/my-service/.garden/kubernetes.garden/build/registry-proxy

Error: no Chart.yaml exists in directory "/Sandbox/git/my-service/.garden/kubernetes.garden/build/registry-proxy"
    at makeError (/snapshot/dist/node_modules/execa/lib/error.js:58:11)
    at handlePromise (/snapshot/dist/node_modules/execa/index.js:112:26)
    at process._tickCallback (internal/process/next_tick.js:68:7)
Error Details:
command: >-
  $USER_HOME/.garden/tools/helm/1c6ed8021f49d4ce/darwin-amd64/helm
  --tiller-namespace garden-system --kube-context kubernetes template --name
  garden-registry-proxy --namespace garden-system --values
  /Sandbox/git/my-service/.garden/kubernetes.garden/build/registry-proxy/garden-values.yml
  /Sandbox/git/my-service/.garden/kubernetes.garden/build/registry-proxy
exitCode: 1
exitCodeName: EPERM
stdout: ''
stderr: >-
  Error: no Chart.yaml exists in directory
  "/Sandbox/git/my-service/.garden/kubernetes.garden/build/registry-proxy"
all: >-
  Error: no Chart.yaml exists in directory
  "/Sandbox/git/my-service/.garden/kubernetes.garden/build/registry-proxy"
failed: true
timedOut: false
isCanceled: false
killed: false

Expected behavior

Garden services and all garden system specific namespaces removed (e.g.garden-system and garden-system--metadata).

Reproducible example

Install garden services to a cluster: garden plugins kubernetes cluster-init
Uninstall garden services from the cluster: garden plugins kubernetes uninstall-garden-services

Your environment

Garden Versions Affected:

• 0.10.6
• 0.10.11

kubectl version:
Client Version: version.Info{Major:"1", Minor:"16", GitVersion:"v1.16.0", GitCommit:"2bd9643cee5b3b3a5ecbd3af49d09018f0773c77", GitTreeState:"clean", BuildDate:"2019-09-19T13:57:45Z", GoVersion:"go1.12.9", Compiler:"gc", Platform:"darwin/amd64"}
Server Version: version.Info{Major:"1", Minor:"12+", GitVersion:"v1.12.10-eks-825e5d", GitCommit:"825e5de08cb05714f9b224cd6c47d9514df1d1a7", GitTreeState:"clean", BuildDate:"2019-08-18T03:58:32Z", GoVersion:"go1.12.9", Compiler:"gc", Platform:"linux/amd64"}

[edvald]

At first glance, I can only reproduce this if I delete the .garden directory between running cluster-init and uninstall-garden-services. Otherwise it appears to work as expected.

This is basically an issue with the helm module type, where the delete handler doesn't work if the module hasn't been built. Lowering the priority, but should still be easy and quick to fix.

[garoose]

This is basically an issue with the helm module type, where the delete handler doesn't work if the module hasn't been built. Lowering the priority, but should still be easy and quick to fix.

This is a slightly bigger issue for us. We run our garden tests with CI on a regular interval in a persistent environment where sometimes nothing has changed and the tests don't actually need to run. This means the namespace is created, but none of the services are deployed. Then when we try to delete the garden environment with garden delete env, it fails in this same way. Our workaround is to kubectl delete the namepsace after just in case.