Closed rcgoodfellow closed 11 months ago
Part of the challenge here is that vCPUs will race to declare themselves the source of a triple-fault when any one of them incurs one. This has been written up in illumos#16016. Addressing that issue would solve the case of multiple triple-fault events being queued in the state machine.
It is still not clear why we were unable to make forward progress in processing those events (through two reboots in quick succession) during the event in question.
I have a Helios machine running inside Propolis. Issued a reboot after a pkg update in the guest and things seem to be stuck.
Here is the log from
propolis-server
.