Hi everyone, The box crashed early this morning. I didn't have BREAK_TO_DEBUGGER enabled in the kernel so I couldn't get into the debugger. I fixed that so hopefully I can get that next time.
I haven't seen any vr errors (thanks for telling me how to). Indeed as someone else pointed out, it's likely not related to 'vr' at all. When it last hung the 'top' execution stopped about 15 minutes before the box stopped routing packets; but during the last runs the interrupt "busy %" according to top was only 30-ish percent. I didn't log the interrupt rate when it was crashing; but the ifconfig and dmesg outputs didn't show anything. (Though my logging jobs might have stopped well before the system failed, similar to the shell running top). Grasping at straws here, but when enabling BREAK_TO_DEBUGGER I noticed that the kernel had ZERO_COPY_SOCKETS and TCP_SIGNATURE enabled. I disabled those, "just in case". - ask
