https://rancher.com/ logo
Join Slack
Powered by
This message was deleted.
# rke2
a
This message was deleted.
a
The only one I've come across is specific to vSphere deployments and using the VXLAN encapsulating method. What issues are you experiencing?
m
i am also using a vpshere/ Vcloud enviroment, but specific the problem is that pods on the different nodes are unable to communicate, ping is working, but for example curl requests not. For example from master to worker nodes.
a
Are you using Rancher to deploy the cluster?
m
No we dont use Rancher to deploy the cluster.
we just run the installer script. so as stated in the quickstart
a
Can you change the flannel backend from 
vxlan
to 
host-gw
and see if that helps
m
ehm how could we change that
nevermind we had to look
We changed it, lets test if it works brb
its working!
Thanks for the help
What could be the reason why vlxlan as a back-end is not working anymore on Ubuntu 22.04 in a vmware enviorment?
a
I don’t think that particular problem was ever addressed in Flannel (since it’s a vSphere bugfeature). Whether an environment is affected by this, it depends also on the kernel version of the guest OS. It can usually be solved by disabling VXLAN checksum hardware offloading using ethtool :
Copy code
$ ethtool -K <VM network interface> tx-udp_tnl-segmentation off
 $ ethtool -K <VM network interface> tx-udp_tnl-csum-segmentation off
Or changing the overlay protocol like you did See also: https://github.com/flannel-io/flannel/issues/1279 And https://github.com/projectcalico/calico/issues/4727
m
Great thanks for the very insightfull help
Wouldn't it be advisable to consider filing a bug report with VMware regarding the impact it may have on running Kubernetes clusters on VMware?
214 Views
Open in Slack
PreviousNext
Powered by