vSphere provisioning: DHCP on nodes makes them fail

Rancher 2.x
1

Hi!
I’ve been trying and debug this thing for more than a week now, i have a Rancher 2.6 environment (but it happened even in Rancher 2.5.9) and i want to provision a Cluster using vSphere.

Following the docs everything goes well and the cluster deploys properly.

Issue is, after a hour of two from the deployment, the cluster starts misbehaving and after a while plain fails, because the DHCP lease has gone and one of the master nodes has changed his IP.

DHCP is provided by a PFSense on a “private DMZ network” configured as a VLAN on vSphere, configuration on cloudinit is completely clean and i’m running stock Ubuntu Server 20.04 cloudinit images.

Am i doing something wrong?

Thanks!

2

Hi twin17,

Could you create an issue in GH rancher/rancher? Rancher’s Engineers will follow it up.

3

Hi twin17,

You may need to assign a DHCP reservation to those nodes so they always get the same IP.

4

Hi Jack!
I will create the issue ASAP.

By the way about the reservation, with PFSense i can’t make a reservation inside the DHCP Pool, so after the VM has been provisioned, actually inserting the reservation (which has to be a different ip from the one distributed) will change the IP for the VM and trigger the issue itself.

Also having a look at the video “Kubernetes Master Class - Provision Kubernetes in vSphere with Rancher” which comes from Rancher Labs itself, it is said that actually DHCP is the intended way of working with it (16:47) so i just thought i was missing something, but i will definitely move further discussions to GH.

Thanks!

5

Hi @twin17
Did you ever find a solution to this. I am having similar issues.
IP reservation or settings static IP via the cloud-config would be helpful.
Was there ever a ticket raised?

Cheers.
Marshall.