att-comdev
docs: ovs-vswitchd container configured to use interfaces 'enp11s0f0' and 'enp12s0f0'
Is this a bug report or feature request? (choose one): Documentation feature request
Explanation Compute nodes are with interfaces eth0,eth1,eth2,eth3 ie with biosdevname=0 .Hence ovs-vswitchd container bridges are not configured correctly. Its expecting interface name to be enp11s0f0 and enp12s0f0 kubectl exec ovs-vswitchd-xzqsc -n openstack -- ovs-vsctl list-ports br-physnet1 enp11s0f0 phy-br-physnet1
kubectl exec ovs-vswitchd-xzqsc -n openstack -- ovs-vsctl list-ports br-ex enp12s0f0
Error kubectl exec ovs-vswitchd-xzqsc -n openstack -- ovs-vsctl show |grep No error: "could not open network device enp11s0f0 (No such device)" error: "could not open network device enp12s0f0 (No such device)"
Kubernetes Version kubectl version Client Version: version.Info{Major:"1", Minor:"5", GitVersion:"v1.5.2", GitCommit:"08e099554f3c31f6e6f07b448ab3ed78d0520507", GitTreeState:"clean", BuildDate:"2017-01-12T04:57:25Z", GoVersion:"go1.7.4", Compiler:"gc", Platform:"linux/amd64"} Server Version: version.Info{Major:"1", Minor:"5", GitVersion:"v1.5.4", GitCommit:"7243c69eb523aa4377bce883e7c0dd76b84709a1", GitTreeState:"clean", BuildDate:"2017-03-07T23:34:32Z", GoVersion:"go1.7.4", Compiler:"gc", Platform:"linux/amd64"}
Helm Client and Tiller Versions Client: &version.Version{SemVer:"v2.2.2", GitCommit:"1b330722aafcb8123114ae51f69d1e884a326f3e", GitTreeState:"clean"} Server: &version.Version{SemVer:"v2.2.2", GitCommit:"1b330722aafcb8123114ae51f69d1e884a326f3e", GitTreeState:"clean"}
Development or Deployment Environment?: Deployment
Expected Behavior:
ovs-vsctl should not return no such device
What Actually Happened:
Error in ovs-vsctl list-ports
How to Reproduce the Issue (as minimally as possible):
Add a compute node with interface named eth* ie biosdevname=0
Any Additional Comments:
Add documentation for br-ex and br-physnet1 physical interfaces and expected name for that.
i'm hesitant to call this a "bug". to be clear for any members that pick this up, this is mainly a documentation request as values.yaml are can be overridden. we do want to pick some potentially more interfaces that are more "sane" than enp11s0f0 and enp12s0f0, and document where physnet is coming from.
We should perhaps consider changing these to old style device names e.g. eth0 etc in the defaults when updating the documentation.
i think it goes a bit beyond just documentation though. i really think that the interfaces used currently (enp11s0f0 and enp11s0f0) are terrible examples (they were PoC for our use-case). i think we can edit values.yaml for more sane defaults (eth0 and eth1 or perhaps even eth2). i'm working on the issue, and testing in a lab that actually has these interfaces set up.
a community member @RichWellum has offered to look into and pick this one up. leaving it assigned to myself so he can work this item.
I haven't seen much activity on this from @RichWellum. We can work on tidying this up once we move into OpenStack and decide on some more sane defaults if necessary. We can also make sure the ability to override these to any interface desired is clear (if not clear already). I think we can safely close this issue out. @intlabs @larryrensing thoughts?
@wilkers-steve can we leave issues open, and close PR's that aren't applicable? we'll have an exercise of collecting all the open issues, and creating bugs/blueprints when we get into openstack. if they're all closed, we'll miss them. and we definitely want to capture this work.
I got a bit derailed by the k8's to 1.6 and then the AIO was broken the end of last week. I'll have a few cycles early this week. I'm also fine picking it up post-move to OS, or anything else.
