> I think lxd init should account for the fact that not all interfaces
are 1500 by default.
I agree with this. Even if we can make it work on GCE, the standard
requires that all interfaces in a broadcast domain should have the same
MTU. A mismatch even between the host and the container should cau
Or maybe this is a problem with gvenic not handling this correctly, as
this should `just work` as I understand it after speaking with a few
colleagues (even with the bridge mtu delta)
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
htt
gvenic instances seem to have an mtu of 1460 by default. If you `lxd
init --auto` and then `ip link set mtu 1460 lxdbr0`, then the launched
containers should work. I think lxd init should account for the fact
that not all interfaces are 1500 by default.
--
You received this bug notification becau
** Changed in: linux-gcp (Ubuntu)
Assignee: (unassigned) => Kevin Becker (kevinbecker)
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2073620
Title:
http timeout from lxd container using n4-high
I did an image bisect on Canonical's GCE jammy images and found that the
behavior started between 20240207 and 20240207a, which have a manifest
diff of:
$ diff 20240207-manifest.txt 20240207a-manifest.txt
356,362c356,362
< linux-gcp6.2.0.1019.21~22.04.1
< linux-gcp-6.2-headers-6.2.0-10196
