Different networking behaviour across VMs
I have a few VMs (debian jessi) running on a openstack based cloud. The cloud are built with a few PC running Ubuntu. There is a control node, a networking node and a few compute node.
Each VM has a floating external IP assigned with which it can be accessed from outside. Now from the VM I'll have to access a group of remote machines through a GRE tunnel. Strange things is, different VM has different problem but none of them can ping all of the remote machines. VM A might have problem pinging remote machine B but can ping the other ones. VM B will be able to ping remote machine A but no problem with the rest.
I've tried reducing the MTU on the networking node but it didn't help. Given that the VM behaves differently, this looks like a openstack issue. Any suggestion welcomed.