Network problems after compute node reboot
We rebooted a compute node while running an instance on it. After the reboot the network of this instance did not work. I could not ping the internal address nor the floating IP. Rebooting, Hard resetting... nothing helped. I had to recreate the instance based on the volume. Then it worked again.
What could been the problem?
It is Queens running on Ubuntu 16.04 with networking done by Neutron.
Before rebuilding the instance I would first have tried to just detach and reattach the instance's interface. Did you check if neutron (linuxbridge or ovs, whatever you're using) was up and running on the compute node? Did you check with tcpdump where the packages got lost?
Yes, I tried detaching and reattaching the interface. When I created a new instance, everything worked, so linuxbridge on the compute node must be ok. If that happens again I will check with tcpdump. Anyone any other idea?
When you reattached the interface did you check if the instance really had an interface? With something like vnc console? Sounds like the instance's interface configuration could have got lost. Was it running during compute reboot?