search

aws / aws-app-mesh-controller-for-k8s

A controller to help manage App Mesh resources for a Kubernetes cluster.
Apache License 2.0
187 stars 109 forks source link

Mesh custom resource can't be deleted if not owned #101

Open nckturner opened 4 years ago

nckturner commented 4 years ago

When using a mesh that isn't owned by the controller (we only create the mesh if it doens't already exist) the controller then assumes it owns the mesh, and we cannot delete the mesh object unless we can delete everything inside the mesh.

We should allow the mesh to be used by the controller, but not owned. This means the controller should not try to delete the mesh when the mesh custom resource is deleted.

I propose adding:

apiVersion: appmesh.k8s.aws/v1beta1
kind: Mesh
metadata:
  name: my-mesh
spec:
  ownership: shared
M00nF1sh commented 4 years ago

we can support this via tagging on resources we created. marking as enhancement.

RafalMaleska commented 3 years ago

I'm having the same issues with the Mesh resource when using a multi-cluster k8s setup. Are You planing to work on it?

vam-noths commented 3 years ago

Hi, So I am facing this same exact issue:

Image: 602401143452.dkr.ecr.us-west-2.amazonaws.com/amazon/appmesh-controller:v1.4.0

We have an existing ECS Mesh and we would like to use EKS and deploy services into the existing ECS mesh, however this cannot be done as it tries to delete the existing mesh and recreate it - this is not an enhancement this is a bug, considering the whole point of the mesh is to allow seamless communication between the various AWS platforms / external platforms - if the controller insists in trying to take ownership of the mesh, then this will break the other deployments.

igreg commented 2 years ago

I'm having the same issues with the Mesh resource when using a multi-cluster k8s setup. Are You planing to work on it?

@RafalMaleska Did you manage to work around the issue with appmesh controller being unable to manage resources in a mesh it does not own?

@fawadkhaliq Will this issue ever be fixed or is there is a workaround that's been found/implemented since it was first reported?

BrokenKage commented 4 months ago

I am still facing this issue in 2024. I want to have a shared AppMesh between two clusters that survives clusters being destroyed in a blue/green setup. Even using an existing AppMesh would be useful.