So here is the documentation about how you should manager multiple kubernetes clusters from your workstation. I know that now it is mostly CNCF which guidelines the k8s but still google 'baba' is the father of it (or may be mother or both).
https://kubernetes.io/docs/concepts/configuration/organize-cluster-access-kubeconfig/
I was kind of confused when I read this first, not just about the technicality but when they suggest that we should *merge* or have one file for all. In the world of compartmentalization, seggregation of apps or services, microservices, namespaces where application or service distancing is as much necessary as social distancing during the pandemic (read covic19) to minimize the spread, kubernetes options or suggestions to have them all in one file is simply baffling. The dis advantages of official documented methods is
When I type k8cs in my terminal, I switch to a 'k8s cluster sanjose', When I type k8st 'k8s cluster staging'. you can even label then as
k8a
k8ab
k8ac
k8ad...
k8+2 letter permutation among the 26 alphabets.
With just 3 or 4 keystrokes i can switch between clusters. All kubeconfig files are separate from each other and when a colleague wants access or a kubeconfig file of a particular cluster, I can just pass it on without having to worry about passing on a file which contains access to all clusters. I sometimes also append a command to prints out the cluster-info after it switches to that cluster.
https://kubernetes.io/docs/concepts/configuration/organize-cluster-access-kubeconfig/
I was kind of confused when I read this first, not just about the technicality but when they suggest that we should *merge* or have one file for all. In the world of compartmentalization, seggregation of apps or services, microservices, namespaces where application or service distancing is as much necessary as social distancing during the pandemic (read covic19) to minimize the spread, kubernetes options or suggestions to have them all in one file is simply baffling. The dis advantages of official documented methods is
- It is one file for all, If one file is corrupt then your access to all your clusters is in danger.
- A misconfigured master kuber config can lead to right operations performed on wrong cluster, leading to disaster.
- switching between clusters (contexts) is a loooong command
- Messy
- Hard to keep a track of contexts once the clusters you manage grows
Solution: 20 year old solution. Yes. Linux/Unix has a built in solution.
In my daily operations I use alias.
Take a look at my ~/.bashrc file.
alias kl=kubectl alias skl='sudo kubectl' alias cici='export KUBECONFIG=/home/batman/.kube/configs/ramci' alias cist='export KUBECONFIG=/home/batman/.kube/configs/ramstaging' alias k8ra='export KUBECONFIG=/home/batman/.kube/configs/kube_config_rancher-cluster.yaml' alias k8cs='export KUBECONFIG=/home/batman/.kube/configs/kube_config_k8s-devel-cluster.yaml'
k8a
k8ab
k8ac
k8ad...
k8+2 letter permutation among the 26 alphabets.
With just 3 or 4 keystrokes i can switch between clusters. All kubeconfig files are separate from each other and when a colleague wants access or a kubeconfig file of a particular cluster, I can just pass it on without having to worry about passing on a file which contains access to all clusters. I sometimes also append a command to prints out the cluster-info after it switches to that cluster.
No comments:
Post a Comment