I recently encountered an interview question asking about the maximum number of deployments or pods you can run in a k3s cluster, but I'm still confused. My recollection isn't clear, and I've seen conflicting information even from chatbots. Can anyone clarify the limits or key factors affecting this?
5 Answers
Honestly, I think the question itself is a bit silly. It's a very specific number that can change based on the k8s distribution. Instead, focusing on why there are limits to pods or deployments would be more relevant. Interrogating how to work around these limits is what is really needed in a job context!
You're actually limited by the pod IP address range, and while deployments themselves can be unlimited, pods per node are capped. By default, Kubernetes allocates a /24 chunk of IPs to each node, which explains the 110 limit to avoid running out of available IPs for essential services.
In my experience, it mostly comes down to your CNI setup. The default of 110 is common due to IP table limitations, but different CNIs can alter that significantly. Some like cilium allow way more pods because they handle routing differently.
There’s definitely a cap based on node hardware and configuration. I have m5.large instances which usually limit me to 29 pods, but I've overridden it to 60 for specific workloads. Just make sure to manage IP allocations wisely!
The standard limit is 110 pods per node. You can check this by running `kubectl describe node `. Just remember, this is per node, so if you have multiple nodes, you can multiply that limit by the number of nodes in your cluster!
Good point! Just keep in mind that it's a per node limit, so if you have five nodes, that's 550 pods in total.
Exactly! Kubernetes does need a chunk of IP addresses for other services, so that’s why they keep defaults conservative.