Rolling Reload of NonStop Kernel Processors

Rolling Reload of NonStop Kernel Processors 2/18/04
Hewlett-Packard Development Company L.P. 4 of 7 CSSI Website
3 Rolling reload preparation
This section describes the activities that should be performed prior to starting the rolling
reload.
3.1 Prepare the SPRs to be installed
Code to be updated should be present on the preferred load subvolume (often
$SYSTEM.SYSnn, $SYSTEM.SYSTEM, or other $SYSTEM.Z*) depending on
subsystem installation methods. For the best results, use DSM/SCM Build/Apply.
If FUP DUP must be used instead, there is a good chance the next version of code built
with DSM/SCM will not include the changes, leading to an accidental future downgrade
to the earlier version. If FUP DUP is used, you should prepare a DSM/SCM Build with
the updates for the next “official” update with the new versions of the code. Use of FUP
DUP can prevent a rollback from working, which is why it is not recommended.
If new code is placed on the system without using DSM/SCM, a compatible rollback
strategy must be developed.
3.2 Update Firmware
The relevant firmware products should match for all currently running processor
multifunction (PMF) CRUs. All embedded firmware products (such as boot millicode,
service processor (SP), and controller firmware) must be consistent and must be the same
as the latest revision of that product installed on the system. Use OSM/TSM to verify the
consistency of the firmware releases prior to performing the rolling reload. Fix any
discrepancies by using OSM/TSM before continuing.
Note: A NonStop system operation best practice is to keep boot millicode and SP
firmware consistent and as current as possible. Because SP firmware can be updated
without system disruption, NED recommends keeping the revision of SP firmware within
one revision of the current SPR.
3.3 Verify system health
No OSM/TSM alarms should exist on the internal or external ServerNet fabrics. No disk
path downs, disk media errors, or path switch alarms should exist. Examine and correct
any problems found prior to the rolling reload. Correct any OSM/TSM alarms in
power/environment. SCF STATUS SERVERNET $ZSNET must show that both X and Y
fabric connectivity is available to all installed processors.
All processors except one at most (the first to reload) must be executing the NonStop
Kernel operating system.