 |
» |
|
|
|
This section provides a summary of the software
installation process, a description of the software stack, information
about the Kickstart installation file, and a description of the default
file system layout that is applied to the installation disk. Kickstart Installation Process | |
The XC software installation process begins
on the head node and is based on the Kickstart automated installation
process. The HP XC Kickstart process uses a predefined configuration
file that contains the answers to many of the questions required to
install the base operating system for an HP XC system. The single HP XC DVD
software distribution contains a bootable installation image and an
embedded Kickstart file. When you issue the command to initiate a
Kickstart installation, the installation process finds the Kickstart
file on the DVD, and the Kickstart file then controls the base operating
system installation after you provide a few required responses. After the head node is installed,
it eventually also becomes the golden client, which is the node that represents the configuration from which
all other nodes are replicated. Log Files A record of the HP XC installation process and
the installed HP XC RPMs is stored in the /var/log/postinstall.log file. A record of the base Gnu/Linux system installation is stored
in the /root/install.log file. HP XC Software Stack | |
The HP XC software
stack consists of a combination of HP proprietary, third-party, and
open source software products. Table 2-1 alphabetically lists the software products
that are contained in the HP XC software stack. These software
products are contained on a single DVD and are installed by the HP XC software
installation process. Table 2-1 HP XC Software Stack | Software Product Name | Description |
|---|
HP MPI | HP MPI provides optimized
libraries for message passing designed specifically to make high-performance
use of the system interconnect. HP MPI complies fully with the MPI-1.2
standard. HP MPI also complies with the MPI-2 standard, with restrictions. | HP Scalable Visualization
Array | The HP Scalable Visualization
Array (SVA) provides a visualization component for applications
that require visualization in addition to computation. SVA extends
the HP Cluster Platform hardware configuration with the addition of
visualization nodes, which you can use as a specialized compute node. SVA provides a set of visualization
nodes with advanced graphics cards, HP customized software tools,
a parallel compositing library enhanced for visualization, and third-party
visualization software tools. | | | |  | NOTE: For more information see the SVA documentation set,
which is included on the HP XC Documentation CD. | | | | |
| HP XC System Software Version 3.2.1 | HP XC System Software provides
the installation, configuration, administration, and management tools
to support HP XC systems on HP Cluster Platforms 3000, 4000,
and 6000. | HPC Linux for High Performance Computing | HPC Linux from HP
provides Linux ABI (Application Binary Interface) compatibility, which
provides: The ability to run binary serial codes from compatible
Linux systems Access to community-developed software and access
to a large application catalog
| Linux Virtual
Server (LVS) | LVS provides a system alias that enables user logins to be distributed
across multiple login nodes and single system sign-on for both users
and administrators. | LSF-HPC with SLURM | LSF-HPC, the high
performance computing version of LSF from Platform Computing, has
been integrated with SLURM in response to the growing need for a lightweight,
powerful workload management system that is scalable and can support
parallel, compute-intensive workloads across computing resources. LSF-HPC with SLURM contains the same queuing and scheduling management
as standard LSF, but it is integrated with SLURM to gather information
and manage the compute resources. This integration allows users to
make use of SLURM's simple commands to perform a variety of parallel
tasks within their LSF batch scripts. SLURM also provides administration
personnel a small set of powerful tools to manage the resources of
an HP XC system. | MySQL | MySQL is a third-party
application that creates and modifies the HP XC configuration
and management database (CMDB). | Nagios | Nagios is a system
and network monitoring application. It watches hosts and services
that you specify and alerts you when problems occur or are resolved.
On an HP XC system, Nagios is integrated with SuperMon for
monitoring capabilities. | Parallel
Distributed Shell (pdsh) | The pdsh shell
is a multithreaded remote shell that executes commands on multiple
remote hosts in parallel. | SLURM | SLURM was developed
by Lawrence Livermore National Laboratory and Linux Networks. SLURM
is a resource manager for Linux clusters. It manages the key resource
on an HP XC system: the compute nodes. | Standard LSF | Standard LSF is the industry standard Platform Computing LSF
product used for workload management across clusters of compute resources.
It features comprehensive workload management policies in addition
to simple first-come, first-serve scheduling (fairshare, preemption,
backfill, advance reservation, service-level agreement, and so on).
Standard LSF is suited for jobs that do not have complex parallel
computational needs and is ideal for processing large volumes of serial,
single-process jobs. When you install standard LSF',
you are prompted to install the standard LSF-HPC extensions. These
extensions convert standard LSF into standard LSF-HPC and include
MPI and mpich features. For more information
about where to obtain Platform LSF documentation, see “Supplementary Software Products”. | SuperMon | SuperMon is a highly
scalable, high-speed cluster monitoring system. SuperMon provides
all required node statistics to the Nagios subsystem. System statistics
are tiered, aggregated, and stored in the configuration and management
database. | syslog-ng | The syslog-ng logging tool improves upon
traditional syslog functionality. It supplies more
flexibility to handle logs, adds better filters, and contains a better
forwarding mechanism. | SystemImager | The SystemImager tool synchronizes the
configuration of nodes across the system using image propagation.
This facilitates ease of installation of the initial software and
ease of upgrading software and configuration files. |
See “Related Software Products and Additional Publications” for links and pointers to more information
about the open source and third-party software components that are
integrated into the HP XC core technology. Kickstart Installation File | |
The ks.cfg Kickstart installation
file is provided on the HP XC System Software DVD. Default values provided by HP
reduce the number of answers you have to provide during the installation
session. The Kickstart file differs depending on the cluster platform
architecture, and the appropriate version of the file is included
on the distribution media. Table 2-2 lists the default values defined in the Kickstart
file, regardless of the cluster platform. Table 2-2 Default Values in the ks.cfg File | Item | Default Value |
|---|
| Keyboard type | United States (U.S.) | | Mouse | Generic
three button mouse emulation | | Language used by the installation process | U.S. English | | Language installed on the system | U.S. English | | Desktop manager | GNOME |
You can modify these values after the installation
process is complete by using standard Linux system administration
procedures. Default File System Layout and Disk Partition Sizes | |
Table 2-3 lists
the default file system layout and disk partition sizes that are applied
to the installation disk on the head node. Because the installation
disk size can vary, partition sizes are calculated as a percentage
of total disk size. However, using a fixed percentage of the total
disk size to calculate the size of each disk partition can result
in needlessly large partition sizes when the installation disk is
larger than 36 GB. Thus, limits have been set on partition sizes to
leave space on the disk for other user-defined file systems and partitions.
Use the Linux Disk Druid disk partitioning utility to partition the
remaining disk space according to your needs. During the Kickstart installation procedure, messages
notify you if a calculated disk partition size exceeds the limit and
the maximum partition size is applied instead. Table 2-3 Default Disk Partition Layout on the Head Node | File System Name | Partition Size with /hptc_cluster File System | Partition Size without /hptc_cluster File
System | Maximum Partition Size |
|---|
swap | 6 GB[1] | 6 GB[1] | Not applicable | /boot (AMD Opteron™ and Intel® Xeon™
systems) or /boot/efi (Itanium® systems) | 1% of remaining disk space | 1% of remaining
disk space | 150 MB | / (root) | 29% of remaining disk space | 39% of remaining disk space | 10 GB | /var | 50% of remaining disk space | 60% of remaining disk space | 25 GB | /hptc_cluster (on head node only) | 20%
of remaining disk space | Not applicable | 4 GB[2] |
The HP XC System Software Administration Guide describes the purpose and content
of each file system. The default disk partition layout for client nodes
(that is, nodes other than the head node) is listed in Table 3-7. 4 GB /hptc_cluster Partition Size Limit Might Be Too Small
For Some Hardware Configurations | |
The maximum size of the /hptc_cluster partition created during the Kickstart installation is 4 GB, however,
4 GB might be too small for some hardware configurations. As listed
in Table 2-4, the number
of nodes in the hardware configuration and whether or not you are
configuring improved availability of services determines if 4 GB is
acceptable or is too small. Table 2-4 Criteria for 4 GB /hptc_cluster Partition | Plan to Configure
Services With Improved Availability? | Number
of Nodes | Is 4 GB /hptc_cluster Large Enough? |
|---|
| Yes | 0 to 255 | Yes | | Yes | 256 or more | No | | No | 0 to 416 | Yes | | No | 417 or more | No |
Based on the hardware and system configuration,
if the data in Table 2-4 indicates that 4 GB is too small for the /hptc_cluster partition, use the guidelines in the following sections to create
the appropriate sized partition. The guidelines depend on whether /hptc_cluster is located on an HP StorageWorks Scalable
File Server (SFS) or is created on the local system disk. Determining The Size of /hptc_cluster When It Is Located On
An SFS ServerIf you plan to create the /hptc_cluster an SFS file share, use one of the following calculations to determine
the size of the/hptc_cluster file system: If improved availability
of services will be configured on the system, use this calculation:
(8.8 MB * Number of nodes) + 1.75 GB for additional directories and
files If improved availability
of services will not be configured on the system, use this calculation:
(5.4 MB * Number of nodes) + 1.75 GB for additional directories and
files
Determining The Size of /hptc_cluster When It Is Located On
A Local Disk On The Head NodeFollow this procedure if you want to create the /hptc_cluster partition on the local system disk, and
the default maximum limit of 4 GB is too small: Begin the Kickstart installation
on the head node. Answer no when the following prompt appears because you will manually create
the file system in the next step: Do you want to create the XC global file system
(/hptc_cluster) as a partition on the system disk? (y/n) n |
Continue with the Kickstart installation of the
head node. After the head node is installed,
but before running the cluster_config utility to
configure the system, create and mount an /hptc_cluster partition from the space remaining that is remaining on the local
system disk. Use standard Linux administration
procedures to create an appropriately sized /hptc_cluster partition and modify the /etc/fstab file accordingly. Use the following guidelines to determine the
appropriate size of the /hptc_cluster partition: If improved availability of services will be configured,
use this calculation: (8.8 MB * Number of nodes) + 1.75 GB for additional
directories and files If improved availability of services will not be configured,
use this calculation: (5.4 MB * Number of nodes) + 1.75 GB for additional
directories and files
After you create and mount
the /hptc_cluster partition, continue with the
remainder of the cluster configuration process.
|
Printable version
