SSH to CentOS guest hanging when using Ubuntu 13.04 host
I ran out of ideas to try to understand and fix my issue. Before starting voodoo, banging my head on the wall or driving circles around the building with a bicycle, I wanted to make a summary of my findings and ask the group for some ideas.
Summary: I have a customized CentOS Vagrant box that works perfectly on Mac and Fedora host systems but fail on PCs running Ubuntu 13.04: the SSH connection to the guest fails.
Context: I am using Vagrant 1.2.2, VirtualBox 4.2.12 (from virtualbox.org) , Ubuntu 13.04 (64 bits) as a host and the box is a CentOS 6.3 (64 bits). Vagrant is configured for NATing (default)
Issue description: The "vagrant up" command hangs when it start the box and tries to SSH connect to it for post processing. If I am patient enough, Vagrant exits and tells me that the SSH failed. I can see that the box is correctly started as enabled the VB GUI. If I open an other terminal window and try "ssh localhost -vvv -p 2222 -l vagrant" I see that SSH can connect to port 2222 but waits infinitely for the ssh server to send its part.
What I observed:
the same vagrant box works on Mac and Fedora, even with no network connected to the host (should the box startup have network dependencies). Conclusion: it is very unlikely that the issue is (only) with the box.
The same issue occurs on an other Ubuntu 13.04 host with the same VB and Vagrant version. Conclusion: the issue seems linked to the Ubuntu host.
No notable box startup error. Only "kdump" failed. The VB or Vagrant log files don't talk to me: no errors or warning.
a vanilla CentOS 6-3 (CentOS-6.3-x86_64-minimal.box) works: I can SSH to it, I can ping from the host to the guest and vice versa (what fails with my custom box). Conclusion: the VB/Vagrant config works on this host. No iptables configured on the host or the guest. It is the combination that fails.
"ifconfig" on the box looks OK and identical to the one that's working. SSHD configurations are basically equivalent.
"tcpdump" on the guest shows UDP traffic but no TCP nor SSH related traffic
doing a "sudo service network restart" on the guest has a 30s hi-cup when determining the IP address where on the working guest has no delay.
I think that the port forwarding is not working correctly on the Ubuntu configuration.
Any ideas of what I could look to narrow down the issue ?