Closed fwiesel closed 2 months ago
I've tried to reproduce this, but if I set a custom MTU on some neutron network and then boot a cirros instance in that network, I do see eth0 getting the custom MTU value. Can you add some instructions on how to exactly reproduce your issue? (ideally something that does not involve vmware?) ;)
I suspect you could setup a devstack with a non-standard mtu lower than your standard setup like this: https://docs.openstack.org/devstack/rocky/guides/neutron.html#non-standard-mtu-on-the-physical-network
For example:
[[post-config|/$Q_PLUGIN_CONF_FILE]]
global_physnet_mtu = 1450
Our mtu comes from a "standard" neutron network agent with dnsmasq acting as dhcp server.
I still cannot reproduce this. Can you share the output of openstack network show
for your network and then dhcpcd -U eth0
, ip a
and cirros-status
from your instance?
Okay, I have now a KVM hypervisor and there it looks like it should. I'll close it, until I can reproduce it.
It seems the
10-mtu
is usually included in other distributions (debian based at least). Copied in plain-text it is just these lines.