IP space is exhausted

Question

I have an issue about GKE/GCE.

I'm using GKE under shared VPC(alias ip) and I have 4 machines of 2 node pools. When I try to add more node pools(because I want to have more type of machines), it keeps pending and I switched to the GCE/Instance groups tab, it says the IP space is exhausted. But I only have 4 machines in total, the ip range should be enough for 1024(10.129.16.0/22 + 10.130.16.0/22).

Does each node pool will preserve some ip? Like 512 per pool maybe? So when I tried to create more pool, it fails.

Here is a screenshot:

Solution

Just came across this error myself. The documentation is here: https://cloud.google.com/kubernetes-engine/docs/how-to/alias-ips (search for "creation failed: IP space" on that page)

If you look at the table right below it. The "Pod Cidr" space you outlined above 10.129.11.0/24 is the same as "Recommended Pod address range" (right column) in the table. You must have set the "subnet size for nodes" (left column) for your vm to /24, which would reserve 252 ip's for your one vm.

Subnet size for nodes == Alias Ip Range in the their VM creation interface. When creating a VM, click "Management, security, disks, networking, sole tenancy" -> "Networking" tab -> "Show alias IP ranges".

Set the Alias IP range to /29 and now your VM will only request 4 internal ip's.

Comments

[Daniel t.]

How many pods do you have in your cluster? With alias IPs in GKE, containers are assigned IPs from secondary range. Can you post output of gcloud compute networks describe us-east1-ai-pro-secondary-129-a73f3058f217db64 --region us-east1 --project common-prod as well.

[chux0519]

@Danielt. thank you for your reply, I actually don't have permission, so I have to ask my admin to run this command and I will give feedback as soon as possible.

[Sunny J.]

What region was this in? Are you able to create GCE instances outside of GKE in that subnet? Are you hitting any network quotas? Are you able to reserve IPs in that range on their own? Is this still affecting you?

[chux0519]

Thank you all guys. I just noticed that each node will preserve 256 IP, so in theory I can only have 4 node at the same time. But I can't find any doc about this. I will add a picture to prove this. So, the solution is I should contact to my leader and request more IPs.