Preinstallation Planning

The basic process of installing Solaris remains the same, regardless of the installation method selected. A number of planning tasks must be performed prior to installation. These tasks include:

• Choosing the appropriate installation method from the Web Start Wizard, JumpStart, suninstall , and Live Upgrade.

• Deciding whether or not to upgrade an existing installation or install the operating system cleanly. If your system is currently running Solaris 2.6, 7, or 8, an upgrade can be performed. If your system is running Solaris 2.5.1 or earlier, or if it is not running Solaris at all, you need to perform an initial installation. An upgrade preserves many of the system settings from a previous installation, and generally takes less time to complete than a completely new install. If an upgrade is being performed, the current system should be backed up by using ufsdump or something similar so that it can be restored in the event of an upgrade failure.

• Analyzing your existing hardware devices to determine whether or not Solaris 9 will run on your system without an upgrade. For example, Solaris 8 on SPARC would run with only 64MB RAM, but at least 96MB of RAM is required to run Solaris 9. To perform an upgrade installation, extra RAM would need to be added to an existing Solaris 8 system with only 64MB RAM.

• Determining whether your storage devices have sufficient capacity to install Solaris 9 and all required third-party applications. A complete Solaris 9 installation requires 2.4GB of disk space, if OEM support is included, and 2.3GB if OEM support is excluded. A Developer installation requires at least 1.9GB, while the End User installation requires 1.6GB. In addition, an amount of swap space equivalent to twice your physical memory should be factored into the sum, along with third-party and user disk space requirements. This is not a requirement, but a sound practice.

• Choosing an appropriate installation medium. Possibilities include a JumpStart, CD-ROM, DVD-ROM, or net-based installation from a remotely mounted CD-ROM or DVD-ROM drive. For large organizations, it s often convenient to set up a single network server with an NFS-exported DVD-ROM or CD-ROM drive that is publicly available for mounting. In addition, large organizations might also choose a customized JumpStart installation, which also requires network access to a centralized boot server. Smaller organizations will almost certainly use a CD-ROM or DVD-ROM drive attached to the local system to be installed.

• Gathering all of the necessary system configuration information. This includes the system hostname, IP address, subnet mask, name service type, name server IP address, default router IP address, time zone, locale, and proxy server IP address. These values, and when they are required, will be discussed next .

By undertaking a comprehensive preinstallation review, a successful installation can be assured. In addition to making decisions about the installation type and gathering basic system data, it s important to understand the network context in which the system will operate . The network context can be defined by answering several key questions:

• Will the system be networked? If so, you will need an IP address, subnet mask, and default router (unless the system itself is intended to be a router).

• Will the system use the Dynamic Host Configuration Protocol? If so, you will not need to supply an IP address, as a lease over an IP address will automatically be granted to you at boot time.

• Will the system use IPv6, the newest version of the Internet Protocol?

• Will the system form part of a Kerberos v5 realm, to allow centralized authentication? If so, you will need the name of the realm, the administration server s IP address, and the address of the primary KDC.

• Will the system use the Domain Name Service (DNS)? If so, you will need the IP address of a primary and secondary DNS server, which is authoritative for the local domain.

• Will the system use Network Information Service (NIS) or NIS+? If so, the IP address or the hostname of the local NIS or NIS+ server will need to be supplied.

• Will the system make use of the Lightweight Directory Access Protocol (LDAP) for centralized authentication and authorization? If so, you will need to supply the profile server s IP address or hostname.

• Will the system use a proxy server to access the Internet? If so, the IP address or hostname of the proxy server will be required.

Answers to these questions will be required to completely configure the system during installation.

Disk Space Planning

The question of how much disk space you require to install Solaris 9 can only be answered by examining the purpose of the server. For a SPARC system, with 512MB RAM, a complete installation will require 2.6GB for software and 1024MB for swap, as well as space for user data and applications. Extra disk space must be set aside for special features, such as internationalization, and an estimate needs to be made of the size of print and mail spooling directories which lie under /var . Although the default size of /var is usually small in the installation program, mail and print servers will need to increase this, by allowing for a reasonable allocation of spooling space per user.

Caution  

Since a full /var file system caused by a large print job can affect other tasks such as mail, it s important to overestimate rather than underestimate the size of /var .

In terms of applications, an Oracle database server, for example, will require at least 1 “2GB of disk space, for software packages, mount points, and table data. For a development system with multiple users, a projection based on the maximum quota for each user should be computed. For example, if 50 users are allowed 100MB disk space each, then at least of 5GB of disk space must be available for their exclusive use ”as a rule, if users have quotas imposed on them, they should always be guaranteed access to that space. If data on a server is mission critical, consideration should be given to installing some volume management software, as described in Chapter 21.

In terms of specific layouts, the typical file system layout for a SPARC architecture system follows a set of customary, although not required, disk slice allocations . Slice 0 holds the root partition, while slice 1 is allocated to swap space. For systems with changing virtual memory requirements, it might be better to use a swap file on the file system, rather than allocating an entire slice for swap. Slice 2 often refers to the entire disk, while /export on slice 3 traditionally holds older versions of the operating system, which are used by client systems with lower performance (for example, Classic or LX systems that use the trivial FTP daemon, tftpd, to download their operating system upon boot). These systems may also use slice 4 as exported swap space. Export may also be used for file sharing using the Network File System (NFS). Slice 5 holds the /opt file system, which is the default location under Solaris 9 for local packages installed using the pkgadd command. Under earlier versions of Solaris, the /usr/local file system held local packages, and this convention is still used by many sites. The system package file system /usr is usually located on slice 6, while /export/home usually contains user home directories on slice 7. Again, earlier systems located user home directories under /home , but because this is used by the automounter program in Solaris 9, some contention can be expected.

The typical file system layout for an Intel architecture system also follows a set of customary, although not required, disk slice allocations. Slice 0 again holds the root partition, while slice 1 is also allocated to swap space. Slice 2 continues to refer to the entire disk, while /export on slice 3 again holds older versions of the operating system, which are used by client systems, and slice 4 contains exported swap space for these clients . The local package file system /opt is still located on slice 5, and the system package file system /usr is again located on slice 6. Slice 7 contains the user home directories on /export/home . However, the two extra slices serve very different purposes: boot information for Solaris is located on slice 8, and is known as the boot slice, while slice 9 provides space for alternative disk blocks, and is known as the alternative slice.

Device Names

Among the most challenging aspects of understanding Solaris hardware are the device names and references used by Solaris to manage devices. Solaris uses a very specific set of naming conventions to associate physical devices with instance names on the operating system. In addition, devices can also be referred to by their device name, which is associated with a device file created in the /dev directory after configuration. For example, a hard disk may have the physical device name/pci@1f,0/pci@1,1/ide@3/dad@0,0, which is associated with the device file /dev/dsk/c0t0d0 . The benefit of the more complex Solaris device names and physical device references is that it is easy to interpret the characteristics of each device by looking at its name. For the disk example given above, we can see that the IDE hard drive is located on a PCI bus at target 0. When we view the amount of free disk space on the system, for example, it is easy to identify slices on the same disk by looking at the device name:

 # df -k Filesystem            kbytes    used   avail capacity  Mounted on /proc                      0       0       0     0%    /proc /dev/dsk/c0t0d0s0    1982988  615991 1307508    33%    / fd                         0       0       0     0%    /dev/fd /dev/dsk/c0t0d0s3    1487119  357511 1070124    26%    /usr swap                  182040     416  181624     1%    /tmp 

Here, we can see that /dev/dsk/c0t0d0s0 and /dev/dsk/c0t0d0s3 are slice 0 and slice 3 of the disk /dev/dsk/c0t0d0 . If you re ever unsure of which physical disk is associated with a specific disk device name, the format command will tell you:

 # format Searching for disks...done AVAILABLE DISK SELECTIONS: 0. c1t3d0 <<SUN2.1G cyl 2733 alt 2 hd 19 sec 80>>           /pci@1f,0/pci@1/scsi@1/sd@3,0 

Here, we can see that physical device /pci@1f,0/pci@1/scsi@1/sd@3,0 is matched with the disk device /dev/dsk/c1t3d0 . In addition, a list of mappings between physical devices to instance names is always kept in the /etc/path_to_inst file. More information on device naming conventions can be found in Chapter 18.

SPARC Preinstallation

Prior to installing or upgrading Solaris on a SPARC system, it is suggested that a few basic checks of the system be performed, to obtain data necessary for installation (such as the device name of the boot disk) and to verify that all system components are functional. The three most commonly performed tasks are checking network connectivity, checking the disks that have been detected on the SCSI bus, and reviewing how much memory is installed.

If you are booting over a network, or if your system needs to access a DNS, NIS/NIS+, Kerberos, or LDAP server, and you want support for these services to be installed, your network connection will need to be operational. In order to ensure that packets are being sent and received to your system, you can use the watch-net command:

 ok watch-net Internal Loopback test - succeeded External Loopback test - succeeded Looking for Ethernet packets.