Anton, I did some more investigation, and get the impression that this is related to the bug we encountered during upgrade where the private gateway network is associated to the vpc hence creating a nic on the vpc router that has gets the adress of the gateway. If we entered a bug for that yet, we should associate it with this one.
@Alena, I will enter a small review request for what I think is a fix. Much appriciated if you look at it. It is for 4.3 but should apply to 4.4 and master as well; https://reviews.apache.org/r/20511/ On Fri, Apr 18, 2014 at 11:16 AM, Anton Opgenoort <aopgeno...@schubergphilis.com> wrote: > Hi, > We've hit a variation on an older (fixed) bug regarding VPC private gateways. > The old case which prevented a restart of the VPC router (cloudstack-4193) > was fixed; but this time, we could not spin up a virtual router when the VPC > router was destroyed, unless the private gateway was removed from the VPC > config. > > The use case why we need to destroy the VPC router: we need to move the > routerVM away from the cluster it is running on, since the cluster will be > decomissioned. RouterVM's can't be migrated to other clusters, thus need to > be destroyed to have them created on a new cluster. > This issue prevents a smooth travelling of VPC routers from old to new > clusters. > > Stack traces are in CLOUDSTACK-6448. > > Kind regards, > Anton Opgenoort > -- Daan
