HP-UX vPars and Integrity VM V6.3 Administrator Guide

With vPars and Integrity VM V6.3, the olrad CRA request issued for the physical NIC on VSP
initiates parallel CRA check in each of vPar and VM guests for the guest associated with vswitch.
LAN CRA module in the vPar or VM performs usage analysis and reports any potential impacts
from LAN subsystem perspective. Some of the usage scenarios determined by LAN CRA includes
NIC port configured with VLAN and IP address, connected to network, and so on. For more
information about how a resource analysis is performed on LAN components of a system, see the
white paper Critical Resource Analysis. The usage analysis result from each of the vPar or VM is
consolidated and a cumulative criticality is passed back to the olrad(1M) command.
After the CRA is complete, the olrad(1M) command suspends the card based on the criticality
returned from the CRA. The slot will be powered off and the link state of vswitch goes to LinkDown
and link state of NIC goes to DOWN state. The old NIC can be safely replaced with a newer NIC
card. After the card is replaced using the olrad R command, the link state of the vswitch and
the vNICs associated with it, gets transitioned to UP state.
NOTE: When Serviceguard is configured as a package, guest LAN interfaces used for exchanging
Serviceguard heartbeat packets are backed by a physical NIC on the VSP. A CRA operation on
such NICs may report that the card is DATA CRITICAL. This is consistent with the operation of CRA
host systems using Serviceguard in a non-virtualized environment.
11.4.5 AVIO Storage devices
For non-NPIV based AVIO storage devices (referred to as legacy AVIO storage), unlike the NPIV
devices, the multi-pathing capabilities reside on the VSP and each device has only a single path
within a vPar or VM guest. For example, for a disk backing store, the multiple paths seen on the
VSP map to a single path seen within the vPar or VM guest.
Prior to vPars and Integrity VM V6.3, a storage I/O card in an OLR capable PCI slot on the VSP
can be replaced without bringing down active vPars or VM guests configured with backing stores
having paths configured through the card being replaced. This can be done as long as there is at
least one unaffected path in the VSP to the vPar or VM guest backing stores that are in use. In
addition, you must not have impacted devices on the VSP being critical to the operation of the VSP
itself. The olrad command when run on the VSP cannot distinguish between a system critical
usage and a data critical usage of the vPar or VM backing stores.
With vPars and Integrity VM V6.3, the olrad command when run on the VSP, can distinguish
between a guest boot device and a data device. With V6.3, a storage I/O card on the VSP cannot
be replaced without bringing down active vPars or VM guests booted with an AVIO backing store
if the I/O card impacts the only available path to the boot device of the vPar or VM guests.
Unlike in the case of NPIV devices, the CRA in the case of legacy AVIO devices only consider the
backing store on which the vPar or VM guest is currently booted, as a system critical resource. In
some cases, the current boot device may be different from the primary or alternate boot device
configured for the vPar or VM guest.
NOTE: With legacy AVIO devices, the CRA will not consider the impact to the primary boot (if
different from the current boot device) and an alternate boot disk configurations, swap and dump
device configurations, and Serviceguard lock disk configurations as system critical. All of these
will be reported as DATA CRITICAL impact if the only available path will go down when the card
is replaced. Hence, the usage of the olrad command with the -f option to override the DATA
CRITICAL errors must be exercised with utmost caution.
When the CRA for legacy AVIO storage devices reports the severity for a particular vPar or VM
guest as DATA CRITICAL, the VSP administrator must work with the specific guest administrators
to manually check and ensure that no primary or alternate boot devices, swap or dump devices,
or Serviceguard lock disks are impacted by the unavailability of the card being replaced.
166 PCI OLR support on VSPs