Here are things that still make my live hard on AKS. Although the points from my previous post hold. these ones are the most annoying at that moment (August, AKS 1.10.3 and 1.9.6, West Europe Region):
Two main pain points
These ones make my hair turn white.
- Memory-preserving updates, that might kick in, you will get alerts from your monitoring solution, and k8s will immediately reschedule your pods. Not cool on Saturday 4:00.
- AKS team applying changes to the cluster without informing you, e.g., I woke up one day and I had the RBAC activated on my production kubernetes. I am aware it is managed, but when we are in the early GA phase and still there are some issues, I would prefer to know about it.
I got used to it:
- Support :D, still learning the product. Often you are on your own.
Pro-tip from an AKS engineer: resizing your cluster causes the newest configuration to be applied on the master nodes.
Notice: your k8s workers are updated on restart. Draining a problematic node and restarting it, might solve your problem
It gets in the way and slows us down:
- Recently deleting any pod takes ages
- Slow time the persistence storage is attached to your node
- Sometimes you are not able to delete your pod, so you need to force it
If you have more detailed questions, feel free to contact me. The AKS improved a lot in the last weeks.
Good luck! 🙂
pps. We are hiring Golang Developers and System Engineers 😉 – help us to further develop our API and SaaS for finanse processes automation with Machine Learning – https://short.sg/j/1566212