Benjamin Weaver
2012-Jan-16 14:41 UTC
Ubuntu (Pangolin) HVMs become unpingable when migrated back to box where they were created
I have a 2-node cluster running Debian Squeeze as the Dom0 for each of the nodes, and Precise Pangolin Ubuntu DomUs. If I create an HVM instance on one of the Debian nodes and migrate it to the 2nd node, all is well. But if I then migrate the HVM back to the node on which it had been created, it becomes unavailable on the network. A Ping shows it is unavailable. The console hangs for this HVM, suggesting the HVM is in kernel panic although I cannot verify this. I have guessed that there''s a bad housekeeping problem on the originating node that somehow causes a conflict with the original, now returning HVM, and that this conflict interferes with the netfront driver somehow. Any ideas?