https://rancher.com/ logo
Join Slack
Powered by
This message was deleted.
# k3s
a
This message was deleted.
c
have you tried turning off checksum offload?
h
I'm also observing pod to pod communication issues for http traffic (still with icmp traffic being properly transmitted)
have you tried turning off checksum offload?
Not yet, thanks! tcpdump traces were not showing checksum issues reported in the related issue. I'll try
c
Copy code
ethtool --offload eth0  rx off  tx off
h
Thanks a lot ! This seems to fix the symptom. I'll search for related issues to get more background. You're saving my day !
The related blog I had discarded as the tcpdump checksum wasn't showing the offending trace https://t.du9l.com/2020/03/kubernetes-flannel-udp-packets-dropped-for-wrong-checksum-workaround/ I'll further read this in depth. Thanks again for your help.
c
are you by any chance running on vmware?
from what I can tell its much more common on that platform, due to a bug in the virtual nic driver
h
yes, we are reproducing this issue on vmware and not on our openstack kvm infrastructure.
If this may help others while reading this thread later on, I tried to collect associated upstream bugs suhc as kernel bug in vmware vmxnet3 into https://github.com/orange-cloudfoundry/paas-templates/issues/2062
Open in Slack
PreviousNext
Powered by