|
» |
|
|
|
Processor Terminology Processing resources under vPars, both as input arguments
and command outputs, are described as “CPUs.” For multi-core
processors such as the PA-8800 and dual-core Intel Itanium processors,
the term “CPU” is synonymous with “core.”
The term “processor” refers to the hardware component
that plugs into a processor socket. Therefore a single processor can
have more than one core, and vPars commands will refer to the separate
cores as distinct “CPUs,” each with its own hardware
path. Two vPars terms pre-date multi-core processors,
so they are exceptions to this terminology: “boot processor”, which refers to the
CPU (that is, core) on which the OS kernel of the virtual partition
was booted, and “cell local processor (CLP),” which
refers to a CPU on a specified cell.
For more information on
dual-core processors, see “CPU: Dual-Core Processors”. Hyperthreading Hyperthreading is a new feature supported in HP-UX
11i v3 (11.31) environments on servers with the dual-core Intel Itanium
processors. It provides for executing multiple threads on a single
processor core; each thread is abstracted as a “logical CPU”
(LCPU). In vPars A.05.xx, you can enable and disable hyperthreading
with the vPars Monitor; however, in a mixed HP-UX 11i v2/v3 vPars
environment, any virtual partitions running vPars A.04.xx/11.23 will
not boot unless hyperthreading is disabled. For more information on
hyperthreading, see “CPU: Hyperthreading ON/OFF (HT ON/OFF)” and “Setting Hyperthreading (HT ON/OFF) and cpuconfig Primer”. CPUs are assigned to virtual partitions on a core
basis, and not on a logical CPU (LCPU) basis. asyncdsk driver Many applications, such as databases, use the asyncdsk driver to lock down memory for I/O transfers.
As of this writing, the asyncdsk driver does not
support memory deletion. As a result, if the driver has locked down
any float memory, then that portion of memory cannot be deleted from
a virtual partition. See the most recent version of the HP-UX Virtual Partitions Release Notes for more information. PCI On-Line Addition and Replacement (OLAR) Except for the functions stated below, OL* for PCI slots
works the same on a vPars server as it does on a non-vPars server.
Note that you can execute PCI OL* functions only on the PCI slots
that the virtual partition owns. PCI doorbells (the physical attention button on
the system) are supported beginning with the HP-UX December 2003 HWE
release and vPars A.03.01. (PA-RISC only)
In a vPars system, a reboot of the virtual partition does not power
on a slot that was powered off prior to the reboot. If you wish to
power on the slot, you need to do this manually after the reboot using
the rad command: rad -i slot_id. For information on the use of PCI Card OLAR features
on HP-UX 11i v1 (11.11) see Configuring HP-UX For Peripherals:
HP 9000 Computers at the HP Technical Documentation website: http://docs.hp.com/en/oshpux11i.html#System%20Administration For information on the use of PCI Card OLAR features
on HP-UX 11i v2 (11.23) or HP-UX 11i v3 (11.31) see the Interface Card OL* Support Guide at the HP Technical
Documentation website: http://docs.hp.com/en/ha.html#System%20Administration The PCI OL* error recovery features that are supported
in 11.31 are also supported within a vPars environment. For complete
information on PCI OL* error recovery, see the following documents
available at the HP Technical Documentation website (http://docs.hp.com): PCI Error Recovery Product Note PCI Error Recovery Support Matrix
Support Tools For information on the required version of the
Support Tools package that can run on your vPars server, see the section
on Online Diagnostics in the HP-UX Virtual Partitions Ordering
and Configuration Guide. Prior to STM version A.43.00 (December 2003), in a vPars environment if the LPMC (Low Priority
Machine Check) Monitor (now known as CPU Monitor) of the Support Tools
bundle deactivates a processor, it does not automatically replace
the failing processor with an iCAP (formerly known as iCOD) processor.
The processor replacement must be performed manually using theicod_modify command. For more information, see the Instant Capacity User’s Guide. Beginning with STM version A.43.00, in a vPars environment,
the LPMC monitor automatically replaces a failing processor with an
iCAP processor if an iCAP processor is available. CSTM’s CPU Info Tool With STM version A.43.00, the CSTM's cpu
info tool will show information about the bound CPUs assigned to that
partition and unbound CPUs not assigned to any partition. For unbound
CPUs assigned to other partitions, it shows N/A. With STM version C.48.00 (HWE0505) and later,
the CSTM's cpu info tool will show information about all the
CPUs that are assigned to the current partition. When using CSTM’s info command,
information is shown only for the CPUs that are currently owned by
the virtual partition from which the info command
is run. For CPUs which are not assigned to the virtual partition,
the information is displayed as N/A. ODE Diagnostic and I/O Card Utilities (Integrity Only) ODE diagnostics utilities, such as CEC, CPU, or
MEM, do not operate in vPars mode. Also, the I/O card and diagnostic
utilities, such as FCFUPDATE and IODIAG.efi, do not operate in vPars mode. You must boot the nPartition into nPars mode to operate these utilities. For more information on modes, see “Modes: Switching between nPars and vPars Modes (Integrity Only)” Ignite-UX Making Depots for Ignite-UX: For information
on where to find a “cookbook” for setting up your Ignite-UX
server for use with vPars, see “Setting Up the Ignite-UX Server”. Reading the CPU counts from Ignite-UX (vPars A.03.xx
and earlier): When Ignite-UX reports the Total Number
of CPUs for a partition, it includes unassigned, unbound CPUs in the
count. For information on bound and unbound CPUs, see “CPU: Bound and Unbound”. For example, if you have three virtual partitions, each with
one bound CPU, and two unbound CPUs not assigned to any of the partitions,
this is a total of five CPUs in the server. Ignite-UX will report
three CPUs (one bound and two unbound CPUs) for each partition. The data on unbound CPUs is repeated for each virtual
partition. Therefore, adding up the numbers results in a total of
nine CPUs for the server when there are actually only five physical
CPUs.
(PA-RISC only) The WINSTALL Boot Kernel
Paths with Different Versions of Ignite-UX and the vparboot
-I command The examples in this document use the Ignite-UX
bootable kernel WINSTALL path as: /opt/ignite/boot/WINSTALL
This is the correct path for Ignite-UX versions B.05.xx
and earlier. However, if you are using Ignite-UX version C.06.xx or
later, the WINSTALL path has changed to: /opt/ignite/boot/Rel_B.11.NN/?INSTALL where NN completes the
HP-UX version and where ? is W for PA-RISC systems
For example, if vPars is using HP-UX 11i v2 (11.23)
instances on a PA-RISC server with Ignite-UX version C.06.xx, the
path is: /opt/ignite/boot/Rel_B.11.23/WINSTALL
Thus, if you are using HP-UX 11i v2 (11.23) on
a PA-RISC server and Ignite-UX version B.05.xx or earlier: you should specify /opt/ignite/boot/WINSTALL on the command line (/opt/ignite/boot is
the default); for example # vparboot -p <target_partition> -I <ignite_server>,/opt/ignite/boot/WINSTALL |
Ignite-UX version C.06.xx or later: you
must specify the absolute path /opt/ignite/boot/Rel_B.11.23/WINSTALL on the command line (because /opt/ignite/boot is the default); for example # vparboot -p <target_partition> -I <ignite_server>,/opt/ignite/boot/Rel_B.11.23/WINSTALL |
Ignite-UX Recovery and Expert Recovery Beginning with vPars A.02.03, the creation of make_tape_recovery tapes is supported on vPars-enabled
servers. However, there are limitations: Ignite-UX Recovery via make_net_recovery requires additional steps as noted in “Network and Tape Recovery”. Expert
recovery works as documented in the Ignite-UX manual; however, you
must account for the vPars differences described in “Expert Recovery”. For more information on using tape devices, see
also the Booting, Installing, Recovery, and Sharing in a
vPars Environment from DVD/CDROM/TAPE/Network white paper
available at: http://docs.hp.com/en/vse.html#Virtual%20Partitions Ignite-UX and other Curses Applications On the virtual console, when using applications
that use curses, such as the terminal versions of Ignite-UX and SAM,
do not press Ctrl-A to toggle
to the console display window of another virtual partition while you
are still within the curses application. This is especially applicable
when you are using vparboot -Iand the Ignite-UX
application to install vPars. For more information on curses, see
the curses_intro(3X) manpage. As with most curses applications, if you get a
garbled display, you can press Ctrl-L to refresh
the display. ServiceGuard ServiceGuard is supported with vPars. However,
because ServiceGuard is used to guard against hardware failures as
well as software failures, its functionality will be reduced if a
cluster includes multiple virtual partitions within the same server.
Such configurations are not recommended. See the ServiceGuard documentation
for more information on running ServiceGuard with vPars. UPS (uninterruptible power supply) software UPS hardware communicates with UPS software via
the serial port. By default, a hard partition has only one serial
port. For a hard partition that runs vPars, the serial port can be
owned by at most one virtual partition. Therefore, on the hard partition,
the UPS can communicate with only the virtual partition that owns
the serial port. Alternately, the HP PowerTrust II-MR UPS product
can be configured across virtual partitions using network connections,
providing all the virtual partitions reside on the same network. Processor Sets (vPars A.03.xx and earlier) You cannot specify a hardware path for an unbound CPU. Therefore,
to avoid unintentionally removing unbound CPUs from a non-default
pset, initially create the partition that will be running Processor
Sets using only bound CPUs. Then, when you add or remove an unbound
CPU, the unbound CPU will be added to or removed from only the default
pset. Glance and Openview Performance Agent (MeasureWare) For correct reporting of processor utilization,
you need to run Glance and MeasureWare versions C.03.35. or higher. Real-time clock (RTC) Fixed in A.03.03 and later, A.04.01 and later: The vPars Monitor keeps track of the OS time for
each virtual partition relative to the real-time clock. The OS time
is the time that is changed via the set_parms or date commands. However, you can
change the real-time clock at the BCH prompt
or at the vPars Monitor prompt (MON>). If
you change the real-time clock, you need to run the vPars Monitor
command toddriftreset to reset the drifts relative
to the real-time clock. For information on the vPars Monitor commands,
see “vPars Monitor: Using vPars Monitor Commands”. Booting the machine into standalone mode from
a boot disk which had its OS time ahead of the RTC will advance the
RTC. If the machine is then booted into a vPars environment, the OS
time of all the virtual partitions will be advanced. Administrators
should ensure that the RTC is adjusted accordingly before booting
the machine from standalone mode into a vPars environment and vice
versa. SCSI Initiator ID For vPars A.03.xx and earlier: the SCSI Initiator
ID is the ID of the SCSI controller. Although you can display and
set SCSI parameters for the SCSI controller at the BCH prompt, you
can also set these values on a vPars server from the HP-UX shell of
a virtual partition using the vPars command vparutil. For more information, see the vparutil(1M) manpage. For vPars A.04.xx and later: use the mptconfig command to view or set the SCSI parameters for
Ultra320 host bus adapters (HBAs). For information on the mptconfig command, see the HP A7173A PCI-X Dual
Channel Ultra320 SCSI Host Bus Adapter Support Guide.
For Ultra2/Ultra160 SCSI HBAs, the SCSI parameters can only be set
from the BCH prompt (on PA-RISC) or from the EFI Shell prompt using
EFI applications (on Integrity). For information on setting and confirming
SCSI parameters for Ultra2/Ultra160 HBAs, see the HP A6828A
PCI Ultra160 SCSI Host Bus Adapter: Service and User Guide. System-wide stable storage and the setboot command On a non-vPars server, the setboot command allows you to read from and write to the system-wide stable
storage of non-volatile memory. However, on a vPars server, the setboot command does not affect stable storage. Instead,
it only reads from and writes to the vPars partition database. For more information see “Setboot and System-wide Stable Storage ”. mkboot and LIF files The mkboot command allows you
to write to files in the LIF area on both Integrity and PA-RISC servers;
for example, the AUTO file. While on a vPars server, mkboot can still be used to write to files in the LIF area.
However, the LIF area is not read during the boot of an OS on a virtual
partition. Instead, only the information stored in the vPars partition
database is read. (Note that the files in the LIF area are still read
when the system or nPartition boots). To simulate the effect of an AUTO file for a virtual
partition, use the vPars commands so that the information is saved
in the vPars partition database. For more information, see “The AUTO File on a Virtual Partition”. shutdown and reboot commands In a virtual partition, the shutdown and reboot commands shutdown and reboot a virtual
partition and not the entire nPartition. Also, if a virtual partition is not set for autoboot
using the autoboot attribute (see the vparmodify(1M) manpage), the -r and -R options
of the shutdown or reboot commands
will only shut down the virtual partition; the virtual partition will not reboot. In other words, the virtual partition will
halt when the autoboot attribute is not set. For more information,
see the vparmodify(1M) manpage. For the -R and -r options
of the shutdown and reboot commands,
the virtual partition will not reboot when there is a pending Reboot
for Reconfiguration until all the virtual partitions within the nPartition
have been shutdown and the vPars Monitor has been rebooted; note that -R sets a pending Reboot for Reconfiguration. Also, the
requested reconfiguration will not take place until all the virtual
partitions within the involved nPartition have been shutdown and the
vPars Monitor has been rebooted. For more information, see “Shutting Down or Rebooting a Virtual Partition” and “Shutting Down or Rebooting the nPartition (Or Rebooting the
vPars Monitor)”. ioscan output On a PA-RISC system, the ioscan output for vcn and vcs drivers show a value of NO_HW in the S/W State column. On an Integrity server, these drivers
do not appear in the ioscan output. This is normal. intctl Command The intctl command is an HP-UX
tool that enables management of I/O interrupts among the active CPUs.
It can be installed from the HP-UX Software Pack but should be used
only by advanced administrators for performance tuning. If you are
managing interrupts on vPars systems, see the section “Managing I/O Interrupts”. kernel crash dump analyzer You cannot use a kernel crash dump analyzer on
vPars Monitor dumps because vPars Monitor dumps are structured differently
than kernel dumps. For more information on vPars Monitor dumps, see “vPars Monitor Dump Analysis Tool”. top and other applications that show CPU ID The CPU ID displayed by the top command and other applications may not be indicative of the actual
CPU index in standalone or nPars mode, nor of the actual hardware
path. Within a virtual partition, top sees only
the CPUs assigned to it. Possible top output is
shown below; the CPU index is the left-most column. CPU LOAD USER NICE SYS IDLE BLOCK SWAIT INTR SSYS
0 0.01 0.0% 0.0% 0.0% 100.0% 0.0% 0.0% 0.0% 0.0%
1 0.00 0.0% 0.0% 0.0% 100.0% 0.0% 0.0% 0.0% 0.0%
2 0.01 0.0% 0.0% 0.0% 100.0% 0.0% 0.0% 0.0% 0.0%
4 0.01 0.0% 0.0% 0.0% 100.0% 0.0% 0.0% 0.0% 0.0%
7 0.01 0.0% 0.0% 0.0% 100.0% 0.0% 0.0% 0.0% 0.0%
8 0.06 0.0% 0.0% 0.0% 100.0% 0.0% 0.0% 0.0% 0.0%
--- ---- ----- ----- ----- ----- ----- ----- ----- -----
avg 0.02 0.0% 0.0% 0.0% 100.0% 0.0% 0.0% 0.0% 0.0% |
Agile view of Mass Storage The agile view of mass storage introduced in HP-UX
11i v3 (11.31) is supported with vPars. Support for agile addressing vPars configurations
and operations is provided starting with vPars A.05.03. In the vPars
A.05.01 and vPars A.05.02 releases, only the legacy hardware path
format is supported. The vPars A.05.03 release
includes support for the next-generation mass storage subsystem in
HP-UX 11i v3 (including agile addressing). In the next-generation
mass storage subsystem, support has been extended to include vPars
configurations and operations starting with vPars A.05.03. Agile addressing support is only provided for lunpath
hardware paths, not for LUN hardware paths. Use of LUN hardware paths
will result in an error. Agile addressing is supported only for HP-UX 11iv3
(11.31) releases of vPars, starting with vPars A.05.03. It is not
supported by vPars A.05.02 or earlier, vPars A.04.xx on HP-UX 11i v2 (11.23), or vPars A.03.xx on HP-UX 11i v1 (11.11). This restriction also applies to mixed
HP-UX 11i v1/v2/v3 vPars environments. Virtual partitions running
HP-UX 11i v1 or HP-UX 11i v2 do not support agile addressing. In the vPars A.05.01 and
vPars A.05.02 releases, the lunpath hardware path format and LUN hardware
path format are not supported for use on the
vPars command line, and vPars commands will only use legacy hardware
paths in their output. You must continue to use the legacy hardware
path format that existed in previous vPars releases when using the
vPars commands; for HP-UX 11i v3 (11.31), ioscan’s default output will continue to show the legacy format. vPars does not affect the use of the agile view
with non-vPars commands. Wherever the new formats are supported by
other HP-UX commands and tools, you can use these new formats within
the virtual partitions running HP-UX 11i v3 (11.31). vPars does not support disabling legacy mode.
For information on the agile view of mass storage,
including the new hardware paths, device special files, and legacy
mode, see the white paper The Next Generation Mass Storage
Stack in the Network and Systems Management section of http://docs.hp.com, under Storage
Area Management. vMedia Support vPars A.05.03 and HP-UX support read-only use
of optical media via an iLO 2 MP on HP Integrity servers based on
the sx2000 chipset, including rx7640, rx8640, and HP Integrity Superdome
servers. vPars A.05.03 supports
all the standard iLO 2 MP features that are supported by HP-UX, including
the Lights Out Advanced features of vMedia. However, vPars does not
support virtual keyboard, video, and mouse features. Only one virtual partition
at a time can use the Lights Out Advanced features of vMedia. To move the vMedia attachment to a different virtual
partition, the path associated with the vMedia card must be reassigned
to the other virtual partition.
To launch Ignite-UX via a vMedia DVD drive in
vPars mode: Set up a vMedia DVD drive using the Lights Out Advanced
vMedia feature. Insert an HP-UX OE DVD (such as the Virtual Server
OE and Data Center OE) into the vMedia DVD drive. Assign the path associated with the vMedia card to
the target virtual partition. From the vPars Monitor prompt, issue the following vparload command to launch Ignite-UX from the HP-UX OE
DVD in the vMedia drive: vparload -p targetVpar -D fsIndex
vPars and Dynamic nPartition Operations Dynamic nPartitions operations cannot be performed to an nPartition
that is running Virtual Partitions. For more information on dynamic
nPartitions, see the HP-UX 11i v3 Dynamic nPartitions white
paper at: http://docs.hp.com/en/10907/dynamic_nPars_WP.pdf.
|