We should have the host dtraced own all of the state and once something goes wrong in the guest (dtraced dies, --cleanup-state gets passed to dtracedctl, kernel panic, ...) re-transmit it to the guest to restart tracing. This would make the system a lot more fault-tolerant than it is right now.
While this is a pretty "general" issue, a number of smaller, self-contained issues can refer to it as part of the bigger picture.
We should have the host
dtraced
own all of the state and once something goes wrong in the guest (dtraced
dies,--cleanup-state
gets passed todtracedctl
, kernel panic, ...) re-transmit it to the guest to restart tracing. This would make the system a lot more fault-tolerant than it is right now.While this is a pretty "general" issue, a number of smaller, self-contained issues can refer to it as part of the bigger picture.