virtualbox.org

Host: Windows 7 x64
Guest: Ubuntu Server 12.04 (usually run headless) and Ubuntu Desktop 12.04.1 (also previous versions had the same problem), 1 network adapter: NAT w/Intel Pro 1000
Virtualbox: I've tried on all releases from 4.1.10 to 4.1.20

I am having a problem where the outbound NAT networking seems to freeze or stop working periodically from within the guest. Nothing changes on the host and there is no indication of any problems within the guest other than the fact that it just can't connect out to anything. I can't browse websites or ping out from the guest when it happens but nothing seems to show up in the guests logs to indicate a problem.

My setup is one where I have the Ubuntu server set up with apache, nginx, mysql, and php and I use it for doing web development. The problem is that my guest has to pull data from our live database server as well as proxy some data from our live website and when it freezes I just get a gateway timeout message from nginx and nothing works even though the request inbound to my virtual server works okay. After a few refreshes from the browser on my host (pulling from the webservers on the guest), the guest's outbound traffic stops for a time and I just have to wait for it to start working again.

Any ideas or further information I can give? Posts or references to problems similar to this that I have found seem to have been bugs that should be fixed in the current version of VB so I'm not sure what else to try.

Just to confirm, since you don't say so explicitly: I assume the network failure is permanent, not just a temporary glitch?

In the Devices|Network menu item dialog of the running VM window there is a "Cable connected" checkbox. Does toggling this (applying and closing the dialog in between to let the guest see both changes) recover working comms? This is not a suggested workaround, I just want to know if your problem is similar to another I've seen.

I never thought to try that before but when I uncheck and recheck the "Cable connected" box it does seem to recover temporarily but then the problem will return shortly after refreshing the content from my nginx/apache setup 2 or 3 times. I don't know if I would call the problem "permanent" per se because if I wait long enough (maybe 5-10 minutes or so) network functionality seems to work again for a few more content refreshes.

One more side note is that if I'm pinging a host by IP address it seems that I can still ping out even though DNS resolution fails and at least any HTTP traffic from the browser doesn't seem to work.