Disaster Recovery Procedures - HP NetServer AA 4000 Manual

NICs should be on a private switch or a backup network separate from normal user traffic
network for performance reasons. After rebooting HP AA system, check that you can
ping each dedicated IOP IP addresses from a separate outside computer (e.g. network
backup computer). Then check that you can ping each of the CE IP addresses from the
same outside computer.

Disaster recovery procedures

In the event of a disaster or an operator's error, it is possible that the CE OS can be put in an
inoperable state. When this occurs it may be possible to recover the system without restoring
from backup tapes. Below is a set of procedures to assist system administrators to recover
systems with minimum downtime. The procedures must be followed in order listed below.
1. Last Known Good (Control Set)
Some failures of a CE OS to boot (after a reconfiguration) can be recovered from by selecting the
"Last Known Good" configuration. This functionality is exactly the same as the reference machine
(standard NT) case where you strike the space bar on the keyboard when prompted – and prior to the
first log on following the configuration change. Follow below steps:
Boot both IOPs in Marathon Operation Mode and observe if both CE NOS disks are up-to-date
(i.e., both should be green on the Marathon Management GUI). If so, select one of the two tuples
as the "booting tuple." If only one of them is up-to-date, select the tuple containing the the up-to-
date NOS disk as the "booting tuple."
Set the monitor/keyboard switch to enable keyboard on the "booting tuple" side.
Set the monitor switch on SSDL of the "booting tuple" to CE.
Power on only the CE in the "booting tuple".
System will initialize itself. When you receive the message on the CE "Initiating Marathon Boot",
followed by MIC BIOS: CE Boot will be serviced by IOP1 or IOP2 (verify this IOP is indeed in the
"booting tuple.")
After several moments, NTLDR program will starts and you will receive a message to depress
the space bar to load the "Last Known Good" configuration. (Note that the time window allowed
for depressing of space bar can be rather short. Be ready.) Then boots NT with "Last Known
Good" configuration.
Power on the other CE to join the Marathon system.
2. Minor Repair Using An IOP In Off Line Mode
Use an IOP in offline mode to invoke the repair to the CE BOOT disk - then force a mirror copy in the
proper direction as you recover.
Choose an IOP to undertake the repair on the CE BOOT device (cannot be the unfinished target
of a mirror copy)
Reboot the selected IOP to MARATHON OFFLINE MODE (the server already is down/cannot be
rebooted - or you would repair it from the running CE OS).
Replace the driver or delete the service - or otherwise make repairs to the CE BOOT device.
12/08/99 Hewlett-Packard Co. 10