Hi all. I upgraded a cluster to 1.6.1. It seems th...
# harvester
a
Hi all. I upgraded a cluster to 1.6.1. It seems there was a change in where the MAC address of a new VM (in the VirtualMachine resource) is stored. Before the upgrade, it was in
spec.template.spec.domain.devices.interfaces[].macAddress
. Now, it is an annotation called "harvesterhci.io/mac-address". This leads to the experimental add-on "harvester-vm-dhcp-controller" (version 1.6.0, but tested also 1.7.0-dev.0) to fail to create the
VirtualMachineNetworkConfig
resource and thus VMs will not get an IP address from the DHCP server. Is this correct and I just have to wait for a new version of the DHCP controller, or is there a bug with missing macAddress in
VirtualMachine
?
g
a
Didn't see this one, thanks!
g
I had the same issue, hope it will be fixed soon.
a
yes, me too. For the time being, k8snetworkplumbingwg/kubemacpool seems to work for me.
a
I'm using this for a workaround...can't believe they closed that BUG and didn't backport the fix to v1.6.0 or v.1.6.1 helm chart.
Copy code
helm upgrade --install harvester-vm-dhcp-controller harvester/harvester-vm-dhcp-controller \
  -n harvester-system --create-namespace \
  --version 1.6.0 \
  --set image.tag=v1.7.0-dev.1 \
  --set agent.image.tag=v1.7.0-dev.1 \
  --set webhook.image.tag=v1.7.0-dev.1
🙌 1
Any reason the patch wasn't back ported to v1.6.0 or v.1.6.1 for the helm chart before closing the BUG? I'm asking as this could impact anyone trying to test the functionality of this capability on v1.6.1 by using it for RKE2 worker nodes (VMs) vs the other fix of assigning static MAC addresses.