NonStop NS2400 Series Planning Guide

When OSM detects that one power rail is running on UPS and the other power rail has lost power,
it logs an event indicating the beginning of the configured ride-through time period. OSM monitors
whether AC power returns before the ride-through period ends, and:
• If AC power is restored before the ride-through period ends, the ride-through countdown
terminates and OSM does not take further steps to prepare for an outage.
• If AC power is not restored before the ride-through period ends, OSM broadcasts a
PFAIL_SHOUT message to all processors (the processor running OSM being the last one in
the queue) to shut down the system's ServerNet routers and processors in a fashion designed
to allow disk writes for items that are in transit through controllers and disks to complete.
NOTE: Do not turn off the UPS as soon as the NonStop OS is down. The UPS continues to supply
power until that supply is exhausted, and that time needs to be long enough for disk controllers
and disks to complete disk writes.
If a user-supplied rack-mounted UPS or a site UPS is used rather than the HP-supported UPS models
mentioned above, the system is not notified of the power outage. The user is responsible for
detecting power transients and outages and developing the appropriate actions, which might
include a ride-through time based on the capacity of the site UPS and the power demands made
on that UPS.
The UPS and ERMs installed in modular cabinets do not support any devices that are external to
the cabinets. External devices can include tape drives, external disk drives, LAN routers, and
SWAN concentrators. Any external peripheral devices that do not have UPS support will fail
immediately at the onset of a power failure. Plan for UPS support of any external peripheral devices
that must remain operational as system resources. This support can come from a site UPS or
individual units as necessary.
NOTE: OSM does not make dynamic computations based on remaining capacity of the
rack-mounted UPS. The ride-through time is statically configured in SCF for OSM use. For example,
when power comes back before the initiated shutdown, but then fails again shortly afterward, the
UPS has been depleted by some amount and does not last for the ride-through time until it is fully
recharged. OSM does not account for multiple power failures that occur within the recharge time
of the rack-mounted UPS.
This information relates to handling power failures:
• To set the ride-through time for an HP-supported UPS from the manufacturing default setting
to an appropriate value for your system, your service provider can refer to the "Setting the
Ride-Through Time and Configuring for Maximized Runtime" procedure in the NonStop
NS2200 Series and NS2400 Series Hardware Installation Manual.
• To set ride-through time using SCF, see the SCF Reference Manual for the Kernel Subsystem.
• For the TACL SETTIME command, see the TACL Reference Manual.
• To set system time programmatically, see the Guardian Procedure Calls Reference Manual.
Considerations for Ride-Through Time Configuration
IMPORTANT: You must change the ride-through time for an HP-supported UPS from the
manufacturing default setting to an appropriate value for your system. During installation of a
NonStop NS2400 series system or HP UPS, your service provider can refer to the "Setting the
Ride-Through Time and Configuring for Maximized Runtime" procedure in the NonStop NS2200
Series and NS2400 Series Hardware Installation Manual for these instructions.
The goal in configuring the ride-through time is to allow the maximum time for power to be restored
while at the same time allowing sufficient time for completion of disk writes for IOs that passed to
the disk controllers before ServerNet was shut down. Allowing enough time for sufficient completion
of these tasks allows for a relatively clean shutdown from which TMF recovery is less time-consuming
72 System Installation Specifications for NonStop NS2400 Series Systems