a question about high availability

[mertcangokgoz]

I've got a 3-master Kubernetes cluster with no public IPs. Yesterday I noticed something if access to master 1 goes down, all access across the entire cluster gets cut off. NodePort connections and stuff. What could be causing this?

[vitobotta]

Hi, Can you share more details, including the config file (remove the token)?

[mertcangokgoz]

Yeah it's already like that for internet access, but for example, the first master went down somehow and the IP address is timing out.

But traffic isn't coming back through the remaining 2 nodes at all. Even if etcd does an election, the situation doesn't change, I can't figure it out.

[vitobotta]

Which IP address did you select for the gateway? Could it be you selected the master IP address? Try changing it if that's the case

[mertcangokgoz]

I don't think there's an issue with the IP address I picked, because access is available on the private IPs and there's no outage or timeout.

There's no problem with internet access, no problem with node access, it's just that when master1 goes down somehow, access to the other masters gets cut off. I set up a new cluster and tried the same thing there, and it happened there too. Really weird situation. It's either related to the version or some config is breaking it, but I can't figure it out. I'm looking into it and don't know if anyone's come across this before

[vitobotta]

Not sure what could be the problem unfortunately as I haven't tested much with public IPs disabled.

[mertcangokgoz]

I think there are quite a few people who set up and use it with private IP, and doing tests on this might be a good idea. Someone else might've experienced or is experiencing the same thing I'm going through right now I'll look into it in detail — when I find a solution or figure out what's causing the issue, I'll definitely post it here.

Thanks