The following are required to install ACE software:
You must supply the following information:
For each node:
Hostname
Principle IP address
Gateway
For every interface:
IP address
Netmask
Broadcast
Device
Network address
How you do this depends upon your base OS:
For Red Hat, use the control-panel(8) or linuxconf(8) utility. You can also directly edit files in the /etc/sysconfig/network directory.
For SuSE, use the yast(8) utility. You can also directly edit the /etc/rc.config file; if you do this, execute the /sbin/SuSEconfig command upon completion.
The following section describe modifications that must be made regardless of base OS, and an example of the specific changes that must be made for the Red Hat OS.
Modify the following standard configuration files on each node in the cluster.
Enter the following information in the /etc/conf.modules file:
alias device_name_of_ethernet_card network_driver |
For example:
alias eth0 eeprol00 |
The /etc/hosts file on each node must contain at least the IP address and name of the local node. If you do not use a network information service (NIS), it must also contain the IP address and name of each node in the cluster. (For more information about NIS, see the ypbind(8) man page.)
For example, if not using NIS:
On node node01:
#/etc/hosts for node01 192.168.0.1 node01 node01.acme.com 192.168.0.2 node02 node2 node02.acme.com
On node node02:
#/etc/hosts for node02 192.168.0.2 node02 node02.acme.com 192.168.0.1 node01 node1 node01.acme.com
For more information, see the Linux Networking-HOWTO:
http://techpubs.engr.sgi.com/library/dynaweb_docs/linux/usr/HOWTO/Net-HOWTO.html |
To enable users other than root to use MPICH and PBS, you must allow rsh(1) and rcp(1) access to all nodes in the cluster. There are two methods to achieve this:
Enter the hostnames of each node in the cluster in the /etc/hosts.equiv file for each node in the cluster. (The hostname is the output of the hostname(1) command.) You can choose to perform this action within the installace script.
Have each user configure the $HOME/.rhosts file on all nodes in the cluster to allow access from all nodes in the cluster. Ensure that the file is only readable and writable by that user for security purposes.
The files listed in this section pertain to the Red Hat base OS. Other systems will have different files that must be modified.
Enter the fully qualified hostname of the node in the /etc/HOSTNAME file.
For example:
On node node01:
#/etc/HOSTNAME node01.acme.com
On node node02:
#/etc/HOSTNAME node02.acme.com
Enter the IP address of each name server to be searched in the /etc/resolv.conf file.
search domainname.type nameserver node_IP_address |
For example:
#/etc/resolv.conf search acme.com nameserver 192.168.0.31 nameserver 192.168.0.2 nameserver 192.168.0.1 |
For more information, see the resolver(8) man page.
To ensure that root will have rsh(1) access, remove or rename the /etc/securetty file if it exists.
Enter the following information in the /etc/sysconfig/network file:
 | Note: The HOSTNAME value here overrides the value in /etc/HOSTNAME. |
#/etc/sysconfig/network HOSTNAME=hostname DOMAINNAME=domainname.type GATEWAY=gateway_IP_address_for_cluster GATEWAYDEV=interface_card_name_(normally_eth0) NISDOMAIN=NIS_name |
| Note: A gateway is a node on which the routed(8) daemon runs with the -s option to supply routing information. |
For example:
On node node01:
#/etc/sysconfig/network HOSTNAME=node01.acme.com DOMAINNAME=acme.com GATEWAY=192.168.0.1 GATEWAYDEV=eth0 NISDOMAIN="acmepark"
On node node02:
#/etc/sysconfig/network HOSTNAME=node02.acme.com DOMAINNAME=acme.com GATEWAY=192.168.0.2 GATEWAYDEV=eth0 NISDOMAIN="acmepark"
Enter the following in the /etc/sysconfig/network-scripts/ifcfg-eth0 file:
DEVICE=device_name_(normally_eth0) IPADDR=IP_address_of_this_node NETMASK=netmask_to_reserve_for_subdividing NETWORK=IP_address_of_the_network BROADCAST=IP_to_represent_broadcasts ONBOOT=yes_(start_on_boot_process)|no |
| Note: The netmask is used to interpret and define the network portion (subnets included) of the Internet address. This mask normally contains the bits corresponding to the standard network part as well as the portion of the host part that has been assigned to subnetworks. |
For example:
On node node01:
#/etc/sysconfig/network-scripts/ifcfg-eth0 DEVICE=eth0 IPADDR=192.168.0.1 NETMASK=255.255.255.0 NETWORK=192.168.0.3 BROADCAST=192.168.0.4 ONBOOT=yes
On node node02:
#/etc/sysconfig/network-scripts/ifcfg-eth0 DEVICE=eth0 IPADDR=192.168.0.2 NETMASK=255.255.255.0 NETWORK=192.168.0.3 BROADCAST=192.168.0.4 ONBOOT=yes
For more information, see the ifconfig(8) man page.
The installace program requires root to use the rsh(1), rcp(1), and rsync(1) commands from the machine used to install all nodes in the cluster.
The installace script requires that the root user has rsh(1) and rcp(1) access to all nodes in the cluster. This is required for installing the RPM Package Manager packages on the remote node and for verifying system configurations.
To enable users other than root to use MPICH and PBS, you must allow rsh(1) and rcp(1) access to all nodes in the cluster:
For MPICH, the normal process startup mechanism for an ethernet device on networks is rsh. You can select alternatives for MPICH, but doing so will require a source rebuild. For more information, see the Installation Guide for mpich, a Portable Implementation of MPI.
PBS uses rcp to stage input and output files to and from remote destinations. Like MPICH, the user must have rsh permissions on all nodes in the cluster. For more information, see the PBS Administrators Guide.
The pcp, pcp-ace, and pcp-pro packages do not require any special access privileges because they use their own TCP/IP transport services and do not use rsh(1) or rlogin(1).
In order to use the full capabilities of the ACE packages, you must install additional packages from the base OS release. For information about this process, see the base OS release documentation.
Table 3-1 lists the requirements.
Table 3-1. Base OS Packages Required for ACE Packages
ACE Package | Base OS Package Required |
|---|---|
gm | bash, sh |
gm-mpich | csh, ksh, rsh, sh |
lconsole | gtk, sh |
mpich | csh, ksh, rsh, sh |
pbs (known as pbs-server in installace) | sh, tcl, wish |
pbs-mom (known as pbs-client in installace) | sh |
pcp | sh, snmp |
pcp-ace | sh, mpi, textutils (version 2.0e or greater) |
pcp-pro | libstdc++, sh XFree86 |
sgi-ace-tools | sh |
sgi-acedocs | sh |
For example, suppose you want to use the ethernet option to installace to install the mpich , pbs-client, pbs-server, pcp, pcp-pro, pcp-ace, and lconsole set of packages. (See “Install ACE using Interactive Mode” in Chapter 4.) To fully implement these packages, you should install the following from your base OS release:
csh gtk ksh libstdc++ rsh sh tcl wish |
Table 3-1 shows the most important dependencies; it does not list every dependency. For a complete list, use the following command:
# rpm -qR ace_package_name |
For example:
# rpm -qR pbs-mom /bin/sh ld-linux.so.2 libc.so.6 libdl.so.2 /bin/sh libc.so.6(GLIBC_2.0) libc.so.6(GLIBC_2.1) |
Before installing ACE software, verify that you have TCP/IP connectivity and remote access for root.
To verify TCP/IP connectivity, enter the following from the installation machine:
# ping -c1 nodename |
Repeat this for each node in the cluster.
For example, the following output shows that nodes node02 and node01 and have TCP/IP connectivity:
[root@node01 /root]# ping -c1 node01 PING node01 (192.168.0.1) from 192.168.0.1 : 56 data bytes 64 bytes from 192.168.0.1: icmp_seq=0 ttl=255 time=0.1 ms --- node01 ping statistics --- 1 packets transmitted, 1 packets received, 0% packet loss round-trip min/avg/max = 0.1/0.1/0.1 ms [root@node01 /root]# ping -c1 node02 PING node02 (192.168.0.2) from 192.168.0.1 : 56 data bytes 64 bytes from 192.168.0.2: icmp_seq=0 ttl=255 time=0.3 ms --- node02 ping statistics --- 1 packets transmitted, 1 packets received, 0% packet loss round-trip min/avg/max = 0.3/0.3/0.3 ms [root@node01 /root]# |
To verify that the root user can use the rsh(1) command from the machine used for installing to all nodes in the cluster, execute an rsh command such as the following:
# rsh nodename uname |
Repeat this for each node in the cluster.
For example, the following output shows that root on node node04 has rsh capability to nodes node02 and node01:
[root@node04 /root]# rsh node02 uname Linux [root@node04 /root]# rsh node01 uname Linux |