Is this a bug report or feature request? (choose one): feature
Python Version (output of python --version): python3.6
Development or Deployment Environment?: Both
Release Tag or Master: Master
In current implementation, when armada cli is running inside k8s cluster, it looks for incluster-config first and if it fails, then it tries to load kube config. In our use case, that's not right behavior because the cli need to connect to tiller in another k8s cluster. We have a k8s cluster on which armada client pod is running, and we create a new VM that has a new k8s cluster in it. Every time the VM is created, the armada cli needs to connect to tiller in the new k8s cluster. (That tiller is created by "helm init" command.)
Having said that, there are two things that need to be done. Of course, these two things are closely related each other.
Load right config
It should be able to load kube config even though it's running inside a k8s cluster. For example, we can make load_kube_config() have higher priority than load_incluster_config().
Port Forwarding feature to access Tiller
With current implementation, we must specify tiller host and port and that requires the tiller service is exposed as some way like NodePort. That's not always possible. By adding port forwarding feature for tiller service (just like Helm does), this can be solved.
Is this a bug report or feature request? (choose one): feature
Python Version (output of
python --version
): python3.6Development or Deployment Environment?: Both
Release Tag or Master: Master
In current implementation, when armada cli is running inside k8s cluster, it looks for incluster-config first and if it fails, then it tries to load kube config. In our use case, that's not right behavior because the cli need to connect to tiller in another k8s cluster. We have a k8s cluster on which armada client pod is running, and we create a new VM that has a new k8s cluster in it. Every time the VM is created, the armada cli needs to connect to tiller in the new k8s cluster. (That tiller is created by "helm init" command.)
Having said that, there are two things that need to be done. Of course, these two things are closely related each other.
Load right config It should be able to load kube config even though it's running inside a k8s cluster. For example, we can make load_kube_config() have higher priority than load_incluster_config().
Port Forwarding feature to access Tiller With current implementation, we must specify tiller host and port and that requires the tiller service is exposed as some way like NodePort. That's not always possible. By adding port forwarding feature for tiller service (just like Helm does), this can be solved.