We tested a clean method of power cycling a front end. First all user models were killed, followed by the iop. At this point the only things running were the awgtpman process which were restarted by monit. We then IPMI power cycled the frontend (h2susauxb478) and did not affect the daq status of the other running front ends.

I'll write this up as the reboot procedure for frontends in the wiki.

We then simulated a loss of the h2boot server by disconnecting it from the H2FE LAN. The front end medm screens froze up, but did not go "white screen". All medms just froze in a "good" state. The control room workstation of course slowed because they tried to mount /opt/rtcds, and no new medms could be opened because of the file server loss. When h2boot was re-attached to the LAN everything came back alive. Vincent looked at an ISI fast channel and verifed the front end was running and sending data to the DAQ for the whole time.