• packer <=1.5.4. Do not use packer below 1.5.1. For previous packer versions use previous releases from this repository
• Microsoft Hyper-V Server 2016/2019 or Microsoft Windows Server 2016/2019 (not 2012/R2) with Hyper-V role installed as host to build your images
• firewall exceptions for packer http server (look down below)
• [OPTIONAL] Vagrant >= 2.2.5 - for vagrant version of scripts. Boxes (prebuilt) are already available here: https://app.vagrantup.com/marcinbojko
• be aware, for 2016 - VMs are in version 8.0, for 2019 - VMs are in version 9.0. There is no way to reuse higher version in previous operating system. If you need v8.0 - build and use only VHDX.

choco install packer --version=1.5.5 -y

Remove-NetFirewallRule -DisplayName "Packer_http_server" -Verbose
New-NetFirewallRule -DisplayName "Packer_http_server" -Direction Inbound -Action Allow -Protocol TCP -LocalPort 8000-9000

• proper VLAN ID (possible passing as variable -var 'vlan_id=0' ). Look to your build server NIC setings.
• proper Hyper-V Virtual Switch name (access to Internet will be required) (possible passing as variable -var 'switch_name=vSwitch'). Remember - creation of new switch by packer, instead of reusing existing one can cause lack of Internet access. If it's possible substitute variable with your current switch's name.
• proper URL for ISO images in packer's template (possible passing as variable -var 'iso_url=file.iso' ). Be warned - using your own or different images can fail the build, as for example image index or image name used by your ISO can be different then specified by script. Look at the bottom of this Readme to read how to find or use image index.
• proper checksum type (possible passing as variable -var 'iso_checksum_type=sha256' )
• proper checksum (possible passing as variable -var 'iso_checksum=aaaabbbbbbbcccccccddddd' )

• all available updates will be applied (3 passes)

• latest version of chocolatey

• packages from a list below:

  Package	Version
  puppet-agent	5.5.19
  conemu	latest
  dotnetfx	latest
  sysinternals	latest

• latest Nuget poweshell module

• puppet agent settings will be customized (server=foreman.spcph.local). Please adjust it (/extra/scripts/phase-3.ps1) to suit your needs. Puppet is set to be cleared and stopped after generalize phase

• Repositories:

  Repository	Package	switch
  Epel 7		no
  Zabbix 4.2	zabbix-agent	can be switched off by -z false
  Puppet 5	puppet-agent	can be switched off by -p false
  Webmin (CentOS7)	webmin	can be switched off by setting -w false
  Cockpit (CentOS8)	Cockpit	can be switched off by setting -c false
  -	scvmmagent	can be switched off by setting -h false
  neofetch	neofetch	no

• [Optional] Linux machine with separated disk for docker

• [Optional] Linux machine for vagrant

Be aware, turning off latest System Center Virtual Machine Agent will cause System Center fail to deploy machines

• adjust /files/provision.sh to modify package's versions/servers.
• change "provision_script_options" variable to:
  • -p (true/false) - switch Install Puppet on/off
  • -w (true/false) - switch Install Webmin on/off (CentOS7 only)
  • -h (true/false) - switch Install Hyper-V integration services on/off
  • -u (true/false) - switch yum update all on/off (usable when creating previous than latest version of OS)
  • -z (true/false) - switch Zabbix-agent installation
  • -c (true/false) - switch Cockpit installation (CentOS8 only) Example:

"provision_script_options": "-p false -u true -w true -h false -z false"

• prepare_neofetch.sh - default banner during after the login - change required fields you'd like to see in provision.sh

Run hv_win2019_std.ps1 (Windows)

For Generation 2 prepare secondary.iso with folder structure:

• ./extra/files/gen2-2019/std/Autounattend.xml => /Autounattend.xml
• ./extra/scripts/hyper-v/bootstrap.ps1 => /bootstrap.ps1

This template uses this image name in Autounattendes.xml. If youre using different ISO you'll have to adjust that part in proper file and rebuild secondary.iso image.

<InstallFrom>
    <MetaData wcm:action="add">
        <Key>/IMAGE/NAME </Key>
        <Value>Windows Server 2019 SERVERSTANDARD</Value>
    </MetaData>
</InstallFrom>

Run hv_win2019_dc.ps1 (Windows)

For Generation 2 prepare secondary.iso with folder structure:

• ./extra/files/gen2-2019/dc/Autounattend.xml => /Autounattend.xml
• ./extra/scripts/hyper-v/bootstrap.ps1 => /bootstrap.ps1

This template uses this image name in Autounattendes.xml. If youre using different ISO you'll have to adjust that part in proper file and rebuild secondary.iso image.

<InstallFrom>
    <MetaData wcm:action="add">
        <Key>/IMAGE/NAME </Key>
        <Value>Windows Server 2019 SERVERDATACENTER</Value>
    </MetaData>
</InstallFrom>

Run hv_win2016_std.ps1 (Windows)

For Generation 2 prepare secondary.iso with folder structure:

• ./extra/files/gen2-2016/Autounattend.xml => /Autounattend.xml
• ./extra/scripts/hyper-v/bootstrap.ps1 => /bootstrap.ps1

This template uses this image name in Autounattendes.xml. If youre using different ISO you'll have to adjust that part in proper file and rebuild secondary.iso image.

<InstallFrom>
    <MetaData wcm:action="add">
        <Key>/IMAGE/NAME </Key>
        <Value>Windows Server 2016 SERVERSTANDARD</Value>
    </MetaData>
</InstallFrom>

If you need changes For - prepare secondary1903.iso with folder structure:

• ./extra/files/gen2-1903/Autounattend.xml => /Autounattend.xml
• ./extra/scripts/hyper-v/bootstrap.ps1 => /bootstrap.ps1

Run hv_winserver_1903.ps1

If you need changes For - prepare secondary1909.iso with folder structure:

• ./extra/files/gen2-1909/Autounattend.xml => /Autounattend.xml
• ./extra/scripts/hyper-v/bootstrap.ps1 => /bootstrap.ps1

Run hv_winserver_1909.ps1

Run hv_centos81.ps1

• if required change switch_name parameter to switch's name you're using. In most situations packer manages it fine but there were a cases when it created new 'internal' switches without access to Internet. By design this setup will fail to download and apply updates.
• if needed - change iso_url variable to a proper iso name
• packer generates v8 machine configuration files (Windows 2016/Hyper-V 2016 as host) and v9 for Windows Server 2019/Windows 10 1809
• credentials for Windows machines: Administrator/password (removed after sysprep)
• credentials for Linux machines: root/password
• for Windows based machines adjust your settings in ./scripts/phase-2.ps1
• for Linux based machines adjust your settings in ./files/gen2-centos/provision.sh and ./files/gen2-centos/puppet.conf

Experimental support for vagrant machines hv_centos81_vagrant.ps1

• if required change switch_name parameter to switch's name you're using. In most situations packer manages it fine but there were a cases when it created new 'internal' switches without access to Internet. By design this setup will fail to download and apply updates.
• if needed - change iso_url variable to a proper iso name
• packer generates v8 machine configuration files (Windows 2016/Hyper-V 2016 as host) and v9 for Windows Server 2019/Windows 10 1809
• credentials for Windows machines: Administrator/password (removed after sysprep)
• credentials for Linux machines: root/password
• for Windows based machines adjust your settings in ./scripts/phase-2.ps1
• for Linux based machines adjust your settings in ./files/gen2-centos/provision.sh and ./files/gen2-centos/puppet.conf
• no docker repo will be added and no docker-related packages will be installed - this build creates and mount separated volume (size specified by variable) for docker

Run hv_centos78_docker.ps1

Run hv_centos78_docker.ps1

Run hv_centos77_docker.ps1

Run hv_centos77_docker.ps1

Experimental support for vagrant machines hv_centos78_vagrant.ps1 for CentOS 7.8 Experimental support for vagrant machines hv_centos77_vagrant.ps1 for CentOS 7.7

Please create an issue in github. There is slim chance I'll find the time to be your personal helpdesk ;)

Sure. If I can ask - create your PR in smaller sizes, this is repo used for my work, so smaller changes - bigger chances to succeed.

During the deployment secure keys are stored in *.vmcx file and are separated from *.vhdx file. To countermeasure it - there is added extra step in a form of (/usr/local/bin/uefi.sh) script that will check for existence of CentOS folder in EFI and will add extra entry in UEFI. In manual setup you can run it as a part of your deploy. In SCVMM deployment I'd recommend using RunOnce feature.

hashicorp/packer#7278 (comment)

Fixed in version 1.4.4. Do not use lower versions

Fixed in version 1.4.4. Do not use lower versions No resolution so far, template needs to be changed to pass real IP address, or there should be connection between these addresses. Limiting these, end with timeout errors.**

hashicorp/packer#6733

hashicorp/packer#5023 Will be fixed in 1.4.x revision

You can use number. If you have 4 images on the list of choice - use ImageIndex with proper Value

<ImageInstall>
    <OSImage>
        <InstallFrom>
            <MetaData wcm:action="add">
                <Key>/IMAGE/INDEX </Key>
                <Value>2</Value>
            </MetaData>
        </InstallFrom>
        <InstallTo>
            <DiskID>0</DiskID>
            <PartitionID>2</PartitionID>
        </InstallTo>
    </OSImage>
</ImageInstall>

Increase variable update_timeout in ./variables/*.json file - this will create longer pauses between stages, allowing cycles to complete before jumping to another one.

I wish. In short - Windows. These builds should be done with minimum effort (Hyper-V role is enough). Building custom ansible station with lots of checks right now fails in my tryouts.

• Marcin Bojko - marcin(at)bojko.com.pl
• https://marcinbojko.dev/

Work based on https://github.com/jacqinthebox/packer-templates.git