3.2. User Partitioning

Rocks supports user-specified partitioning on Solaris. User-specified partitioning on Solaris is very similar to user-specified partitioning on Linux. We use the concept of programmatic partitioning to specify partitioning on Solaris.

On Solaris, the user specified partitions should be specified in /tmp/user_part_info. This file must be created in the <pre> section of the jumpstart XML file. The contents of this file must conform to the jumpstart partitioning syntax.

Example 3-1. User-specific Partitioning

This example describes how to create a 90 GB root zfs pool on a compute node. There are a few assumptions made in this example. The root device is c1t0d0, and disk is large enough to accomadate a 90 GB root pool, 2 GB swap, and a 1 GB dump.

  1. In /export/rocks/install/site-profiles/5.4.3/nodes/ create the extend-compute.xml file.

    <?xml version="1.0" standalone="no"?>
    
    <jumpstart>
    
    <pre>
    cat &gt; /tmp/user_part_info &lt;&lt; "_feof_"
    usedisk c1t0d0
    fdisk c1t0d0 solaris all
    pool rocks 90G 2G 1G c1t0d0s0
    filesys c1t0d0s3 free unnamed
    _feof_
    </pre>
    
    </jumpstart>
  2. Re-create the solaris distribution.

    # cd /export/rocks/install
    # rocks create soldist
  3. Re-install the node in question, by setting it's bootaction to "install" and rebooting the node.

    # rocks set host boot compute-0-0 action=install
    # ssh compute-0-0 shutdown -g0 -i6 -y

3.2.1. More Information

The syntax for the profile keywords and values can be found at Oracle Solaris 10 9/10 Installation Guide: Custom JumpStart and Advanced Installations.

The default installation uses a ZFS root pool as the root device. Information about how to use this is available here.

Warning

There are certain limitations to using ZFS as the root pool. These limitations are listed here

Warning

It is important to note that the root device chosen must be bootable by the bios. If, for any reason, Solaris cannot determine if device in question is bootable, it will halt at the end of the installation, and the administrator will have to reboot the node manually.