Comment 5 for bug 1912450

On master I'm not able to reproduce this.

I built a multinode devstack:
* devstack0 all-in-one
* devstack0a compute-only

On devstack0a I set the fw driver to iptables_hybrid, added the sleep(5), restarted neutron-agent.

Created a vm:
openstack server create --image cirros-0.5.1-x86_64-disk --flavor m1.tiny --network private --wait --availability-zone :devstack0 vm0

Then moved it back and forth between the computes with 2 resizes:
# 0 -> 0a
openstack server show vm0 -f value -c OS-EXT-SRV-ATTR:hypervisor_hostname
openstack server resize vm0 --flavor cirros256 --wait
openstack server resize confirm vm0
openstack server show vm0 -f value -c OS-EXT-SRV-ATTR:hypervisor_hostname

# 0 <- 0a
openstack server show vm0 -f value -c OS-EXT-SRV-ATTR:hypervisor_hostname
openstack server resize vm0 --flavor m1.tiny --wait
openstack server resize confirm vm0
openstack server show vm0 -f value -c OS-EXT-SRV-ATTR:hypervisor_hostname

After each move from devstack0a to devstack0 I checked for stale flows on br-int in devstack0a by looking at this number (and whether it changed):
sudo ovs-ofctl dump-flows br-int --no-stats | wc -l

However in 10 tries I did not see any stale flows.

Out of curiosity I also did this 10 times with fw driver ovs. Could not find stale flows that way either.

This tells that either this bug is not present on master, or I'm missing something how to reproduce this. Can you reproduce this on newer than rocky versions?