Post-init Scripts

A post-init script is your own bash script that FOS runs at the start of every task, before the task does any work. The script lives on your storage node, not inside the init. So you can change how FOS behaves without building a custom init, and a FOG or FOS update never overwrites your change.

Which customization do you need?

Use the first row that does the job. Each row down is more work to keep current.

You want toUseIt lives in
Change a deployed computer after imaging (drivers, boot entries, config files)Post-download script/images/postdownloadscripts/
Change FOS itself before a task: add a tool, replace a FOS file, set a variable, prepare a diskPost-init script (this page)/images/dev/postinitscripts/
Change FOS before the network or the post-init step starts, or change it on a computer that cannot reach the storage nodeCustom inita custom init file plus the host’s Host Init field

A custom init is a copy of FOS that you must build again for every FOS release. A post-init script applies to whatever FOS release boots. Use a custom init only when a script cannot do the job.

How it works

The installer creates the entry script on each storage node:

/images/dev/postinitscripts/fog.postinit

The installer writes this file only when it is missing. An update does not change it.

At the start of every task, FOS does these steps:

  • It mounts the storage node’s /images/dev/ share over NFS.
  • It sources fog.postinit with the . builtin. Your script runs in the same shell as FOS and sees all of the FOS variables.
  • It unmounts the share, and then it starts the task.

This happens for every task type: deploy, capture, wipe, inventory, Performance Test, and the others. If your script must act on only some tasks, test $type and $mode (see the variables below).

FOS runs from memory. A file that your script writes or replaces inside FOS stays changed until the computer reboots. The next boot starts again from the unchanged init.

Add a script

  1. Put your script in /images/dev/postinitscripts/ on the storage node. For example: /images/dev/postinitscripts/myscript.sh.

  2. Add one line to fog.postinit that sources your script:

    . ${postinitpath}myscript.sh

    ${postinitpath} holds the path of the scripts folder as FOS sees it.

  3. Make the files readable to everyone. FOS reads them over NFS.

If you have more than one storage node, put the same files on each node that runs tasks. FOS reads the scripts from the node that the task uses.

Variables your script can use

These were read in a lab boot of a Performance Test task:

VariableHoldsExample
$macthe host’s primary MAC address08:00:27:71:5e:1f
$hostnamethe host’s name in FOGfos-perftest
$typedown for a deploy, up for a capture, empty for other tasksdown
$modethe task for a non-imaging task, empty for a deploy or captureperftest, wipe, inventory
$osidthe OS ID of the host’s image9
$osfamilythe OS family the server sends. Empty on a server that does not send itwindows
$imgthe image name. It can be empty for a non-imaging taskwin11-lab
$storagethe NFS share that FOS mounted10.255.20.1:/images/dev/
$postinitpaththe scripts folder as FOS sees it/imagesinit/postinitscripts/

To see every variable in your FOS release, add printenv | sort to your script for one boot.

Branch on the operating system

Do not test a list of OS IDs, such as case $osid in 5|6|7|9|10|11). That list breaks each time FOG adds an OS. Test $osClass instead. The post-download page lists each class and its OS IDs.

A post-init script runs before FOS sets $osClass, so set it yourself first. Call determineOS only for a deploy or a capture:

if [[ $type == @(down|up) && -n $osid && $osid != 0 ]]; then
    determineOS "$osid" "$osfamily"
fi
case $osClass in
    win7|winmodern)
        # Windows 7 and later
        ;;
esac

The guard matters. determineOS stops the task with an error for an OS ID that FOS does not know. A host with no image gets osid=0, so a wipe or an inventory task on that host would stop. A deploy or a capture already calls determineOS later in the task, so the guarded call adds no new failure.

$osClass exists in FOS builds released after 2026-11-10 (fos#204). On an older FOS, $osClass stays empty.

Example: replace a FOS file for one boot

This script replaces one FOS file, only on one host, and only for a deploy. The new file is in the scripts folder next to the script:

#!/bin/bash
# /images/dev/postinitscripts/fix-one-host.sh
[[ $mac == 00:11:22:33:44:55 && $type == down ]] || return 0
cp ${postinitpath}myfile /usr/share/fog/lib/myfile

The cp runs before the task starts, so the deploy uses the new file. The same method was tested in the lab: a post-init script replaced /bin/fog.perftest, and the task ran the replacement.

A replaced FOS file does not follow FOS updates

When you replace a whole FOS file, FOS uses your copy even after a FOS update changes the original. An old copy of a file such as funcs.sh can break a newer FOS. Replace the smallest file that does the job. Check your copies again after each FOS update.

Rules that prevent a failed task

  • Use return, not exit. FOS sources your script, so exit stops FOS itself, and the task never starts.
  • Do not leave a command that waits for input. The task waits at the post-init step until someone presses a key.
  • Keep it short. The task does not start until the script finishes.
  • Test on one host first. A test on $mac, as in the example, limits the script to that host.