How to migrate HP-UX workloads between physical and virtual servers easily

Cross-technology moves: a complex problem

Moving a complete OE (OS, applications, and data) to an unlike type of server may require the translation of

meta-information not present in the logical server specification. This meta-information cannot be present in the

logical server definitions since it is either specific to the underlying server where the OE is running or specific to the

kernel of the activated OE. Either keep them “as-is” during the move or translate them to match the target. The

following two examples are typical meta-information:

• Network instance numbers (also known as physical point of attachment [PPAs])

• Extended Firmware Interface (EFI) variables (i.e., boot entries)

During a cross-technology migration, the probability of getting a different hardware path for the network interfaces on

the target is very high. The result is that PPAs in the target kernel can change and the network configuration of the

source becomes inadequate for the target. As an example, if the first network interface (lan0) has 0/0/0/1/0/0/0

for hardware path in a physical system, it may become 0/0/1/0 in a target virtual machine. If no specific action is

performed during the migration, the kernel will bind a PPA different than 0 to 0/0/1/0 when booting on the “V” target.

As a consequence the network configuration file (/etc/rc.config.d/netconf) will not match the operational

PPAs on the target.

Similar to network interface card paths, EFI boot entries are different, depending on the underlying type of server,

although they point to the same bootable device. In a physical environment, we can imagine that the boot device is

reachable at 0/4/0/0/0/0. This location may become 0/0/2/0 after a P2V migration. If the corresponding EFI boot

entry has not been translated accordingly during the move, the operating system of the target virtual machine will

not boot.

The following HP-UX components and features are used to solve the above problems and automate fluid

cross-technology moves:

• HPPortableImage product

• vmVirtProvider WBEM provider product.

• N-Port ID Virtualization (NPIV) technology support

HPPortableImage

The HPPortableImage depot is selected and installed by default in the different HP-UX OEs since March 2012.

For the previous HP-UX versions, you must select it manually from the OE distribution media or Ignite-UX server.

It contains a binary executable (hpuxpitool) launched during shutdown and boot times by sbin/init.d/hpuxpi

when the kernel parameter gio_portable_image is set to 1. During shutdown operations, hpuxpi takes a

snapshot of the network configuration in terms of MAC addresses and PPA associations and saves it in the persistent

Kernel Registry Services (KRS) (see krs(5)) database. During boot time, this saved configuration is restored.

Figure 4 shows the corresponding output on the console.

Note: The hpuxpitool executable is not documented on purpose and should not be used manually. It has been designed to run automatically

or for troubleshooting purposes by HP technicians.