On 15 June 2018 at 05:07, Clark Boylan <cboylan at sapwetik.org> wrote: > On Thu, Jun 14, 2018, at 3:42 AM, Xinliang Liu wrote: >> Hi guys, >> >> The arm64 kolla building job[1] will run containers while building kolla images. >> But we met the mtu mismatched issue[2] between container and vm. That >> makes network accessing very unstable or failed in containers. >> >> I wonder how this issue handled on x86 clouds? And how to fix such issue? > > I believe all of our x86 clouds give us MTUs of 1500 not 1450. I did double check and indeed our linaro instances have MTUs of 1450. > > There are two approaches we can take to fix this. The cloud can be updated to provide MTUs of 1500. I think most clouds do this by allowing their hypervisors to use jumbo frames then the nested overlay networks can use MTUs of 1500 without fragmentation. This would be a cloud side change and may not be possible depending on hardware (this is out of infra's control). > > The other approach is the one detailed in the blog post you linked. Make sure to create any new interfaces with an MTU that is equal to or smaller than your external interface. In this case that would be via docker config as in the blog post. You would update your job to do this. > > Hope this helps, Yes, it's very helpful. will consider the first approach first. Thank you very much:-). Xinliang > Clark > >> >> >> [1] https://review.openstack.org/#/c/557659/ >> >> [2] https://www.variantweb.net/blog/docker-mtu-issues-on-openstack/ >> >> _______________________________________________ >> OpenStack-Infra mailing list >> OpenStack-Infra at lists.openstack.org >> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-infra > > _______________________________________________ > OpenStack-Infra mailing list > OpenStack-Infra at lists.openstack.org > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-infra