© 2004
Canonical Ltd.
•
Terms of use
•
Data privacy
•
Contact Launchpad Support
•
Blog
•
Careers
•
System status
•
35a32c5
(Get the code!)
I'm closing this bug out, because it turns out the problem _was_ actually MTU.
We had a jumbo-frame MTU (of 9216) set on all of the switch ports, and we had an MTU of 9000 set on the OS side. What was missing is, a couple of our port-channels did _not_ have the MTU set. The problem port-channel was on one side of an inter-rack link. If the rack controller and the server being provisioned were in the same rack, then things would be fine (because all of the regular ports had the MTU set correctly). But, if the server being provisioned was in the other rack, then the traffic would go through the inter-rack port channel, and be dropped.
(For the record, we are in the process of spinning up a second rack controller, so each rack will have their own.)
So, for anyone who comes across this problem in the future: If you are using any form of aggregate, make sure you have the MTUs set on the port-channel interfaces!