Posted on: 26/12/2019
A list of sharp edges to watch out for when using Kubernetes. Just to give you an idea about the reliability / quality of Kubernetes and it’s ecosystem, and what to expect!
General
- Multiple terms for the same thing. “resource”, “object”, “kind” and “unit” all refer to the same thing…
- Strange “magic constants” where a blank (
""
) name for an API group refers to the ‘core’ API group… So you get weird error messages like:...cannot get resource "secrets" in API group "" in the namespace "default"
NFS volume mounts
- Do not use the internal DNS https://stackoverflow.com/questions/49274655/kubernetes-cannot-mount-nfs-share-via-dns
- Wrong IP / timeout https://stackoverflow.com/questions/56597634/nfs-volume-mount-times-out-on-kubernetes-with-incorrect-ip
Minikube
- Emojies in terminal output that renders incorrectly on one of the most common Linux terminals https://github.com/kubernetes/minikube/issues/3724
- New “caPitiLizAtion” of command line argument. Results in entire failure of minikube node. Just a generic
"node "minikube" not found
error in logs. https://github.com/kubernetes/minikube/issues/2712
kubectl
- kubectl apply replica count gets reset to a default of 1 if not specified https://github.com/kubernetes/kubernetes/issues/67135
- Events are not sorted by default you need a cryptic
--sort-by='.metadata.creationTimestamp'
parameter - Sort event functionality does not actually sort correctly… https://github.com/kubernetes/kubernetes/issues/29838
Validation
- Error messages can be really really TERRIBLE. Neglected
Container lifecycle
- If your container / pod needs to be killed, no reason will be given. https://github.com/kubernetes/kubernetes/issues/81723
Comments
No comments, yet!
Submit a comment