On Thu, 5 Feb 2004 20:02:13 +0100
"Axel Neumann" <Axel.Neumann@epost.de> wrote:

>
> "David Mathog" <mathog@caltech.edu> wrote:
> > Yesterday I tried attaching a Tripplite SU1500RTXL2U UPS
> > to our V880. This involved plugging in the UPS
> > and then attaching it through a serial line (Sun Serial port
> > ->sun A/B splitter -> B part into DB25 to DB9 cable->UPS rs-232 port).
> Two of the V880 power supplies are plugged
> > into other 120V circuits and one into the UPS (eventually it will be two
> > into the UPS and one separate).
> [.... snip ...]
>
> Hi,
>
> To be on the save side you have to connect at least two of the three power
> cables to an UPS. The V880 can survive the failure of one power circuit but
> it will be very critical if two are failing.

Right, but this was for testing purposes - I wanted the V880 to stay up
even if the software did something stupid and sent an inverter kill
signal to the UPS. When debugging is done it will be 2 lines into the UPS and one through a surge suppressor to a different 120V circuit. That will satisfy the 2 out of 3 rule when the
V880 is running off UPS batteries.

Actually I've already learned that the V880's ttyb _does_ send an
inverter kill (presumably by accident) during the boot sequence, with the unpleasant side effect that the V880 cannot automatically recover from a power failure. Sequence:

1. Line power fails.
2. UPS signals V880 "line failure"
3. V880 shuts down and at the last step sends an inverter kill
to the UPS.
4. Line power returns.
5. UPS starts back up.
6. Power goes to V880, it starts to boot.
7. V880 during boot process toggles RTS high for at least 3.8 seconds
on ttyb.
8. Inverter turns off on UPS.
9. V880 has no power on 2/3 supplies and shuts down.

and it stays that way forever. To reboot the Serial line must be physically unplugged from the UPS. The on button held in on the
UPS for a few seconds. Then a separate utility must be used
to force DTR/RTS high and low (respectively). Then, and only then, is
it safe to plug in the serial line and start the power daemon. Unclear
yet if init.d could be induced to run the serial line
utility early enough in the boot process to avoid killing the UPS.

Previous TrippLites's (omnismart models) only killed the inverter if
there was no line power. So when the voltage leves on the serial line on the attached computer were jumping around during reboot (as they do
on most computers) it made no difference. The SmartOnline
model is apparently "improved" and it kills the inverter
no matter what the state of the power line.

Regards,

David Mathog
mathog@caltech.edu
Manager, Sequence Analysis Facility, Biology Division, Caltech