Author: Aaron

Aaron David Polley is a Canadian-born Musician and MacAdmin based in the Sunshine Coast, Queensland, Australia.

He grew up in a musical family that had a long history of accomplished musicians and songwriters. His own writing ability surfaced at the age of 7 when his first musical arrangement was used in a church service as a congregational song.

As an IT professional of 20+ years, most of his time is now leading 20+ staff at work and a family of 5 at home.

Categories
MacAdmin

SMB Printing on macOS Without Active Directory Binding


The tale of two hostnames

In May 2024, I was assisting a school with setting up a new macOS deployment workflow.

They were previously an all-Windows school and were looking to pilot macOS devices.

Everything went smoothly except for one key issue:

The Mac couldn’t print.

More specifically, the Mac couldn’t print to their SMB printer queues shared/managed by Papercut.

Whilst the official recommendation from Papercut when experiencing SMB issues post PrintNightmare is to use LPD or Mobility Print (IPP/HTTP), sometimes you have to deal with the cards that are dealt 🙂 

https://papercut.com/support/known-issues/?id=PO-522#ng

As described above, the symptom I saw was:

…when printing macOS > Windows over SMB… can result in printers going into a Waiting state indefinitely. You might initially see them go into a Hold for Authentication state - if you click refresh and supply valid credentials, you will end up in the Waiting state.

One of the suggestions in the above thread was to add ?encryption=no to the print queue, which on its own didn’t make any difference.

The baffling part of the issue was that printing WORKED in my initial testing but stopped when I was minting the workflow ready to handover to them.

So what had changed?

I had added a computer naming step to the workflow.

The script used to name the device resembled this:

ScriptRepo/ComputerName-Set-Serial.sh at master · aarondavidpolley/ScriptRepo · GitHub

Changing the computer name in System Settings/Preferences didn’t fix the issue.

Resetting the device, not using the script to name the computer: WORKED!

So why did setting the computer name break it?

If you pay close attention to the script above, macOS has three names:

  • ComputerName
  • HostName
  • LocalHostName

When you use the built-in directory binding plugin (via script or directory utility), the HostName is important to be set before binding as it forms part of the information used in the domain join.

The HostName can be different to the visible name of the computer in Terminal, System Settings, etc., which is why admins have been using scutil for a while to set all three in scripts.

The Jamf binary on a Jamf Pro managed Mac uses the same approach of setting all three names when you use it to set the computer name.

In a world that is ever increasingly free of AD-Bound Macs (to which we rejoice), setting all three computer names is less important and evidently troublesome when trying to print to SMB queues hosted on Windows Server.

Looking at logs, network traffic, and digging around the web, there is evidence that deep in macOS is code that looks for the HostName being set as evidence of an AD-Bound device.

As a result of this confused state of identity, it starts sending RPC signals that AD/Windows Server doesn’t know what to deal with, causing the communication to fail.

Computer: “Hey, I’m this AD-Bound device MAC123, can I print?”

Server: “MAC123 not found in my directory. Go away.”

End of print job attempt.

At the time of this discovery, naming the computer was not an important success criterion.

We skipped naming the device, configured the printer queue install (using a script similar to the one below), that included the ?encryption=no in the queue URI, tested successfully, and off we rode into the sunset.

ScriptRepo/Printers-CUPS-Add-Printer.sh at master · aarondavidpolley/ScriptRepo · GitHub

So why bring this up in September 2025?

I recently visited that same school and helped them revise their macOS deployment workflow. They wanted to expand the pilot to a new set of users and add in any latest tooling/config changes that were appropriate.

Printing came up again.

The papercut environment was largely unchanged, but the queues had stopped working.

After troubleshooting and dusting off old notes (and brain cells), I was able to conclude:

  1. macOS 15.6+ has the same issues as previous versions (including macOS 14, where I had encountered the year prior).
  2. Jamf Setup Manager, which I was now using to set the computer name and therefore using Jamf Pro’s Jamf Binary, was setting all three computer names, repeating the hostname/RPC issue.
  3. Using a script to run scutil --set HostName “” and then rebooting the machine will fix the RPC authentication issue.
  4. The ?encryption=no string now causes the print job to fail and needs to be removed from the printer queue path.
  5. Using a script to only set the LocalHostName and ComputerName via scutil works fine (and is effectively the same as setting the computer name in System Settings).

So in summary:

Symptom: If you can’t print to an SMB printer queue hosted on a Windows Server and it gets stuck waiting or on hold for authentication (even though you’ve provided user auth credentials or have Kerberos tickets).

Check: run scutil --get HostName and see if you get a result.

Fix: if the HostName has been set, run scutil --set HostName “” and then reboot the Mac to fix the RPC authentication issue. 

Hope you find this and it assists your obscure printing issue 😆

Categories
General News

JNUC 2022 Sessions Recap | CompNow Infochino

CompNow’s QLD Engineering Manager, Aaron Polley takes you through his favourite sessions from JNUC 2022 and what the key take aways were overall. He covers everything from Anatomy of Mac Attack to Microsoft Integrations to dealing with IT challenges during war. Learn more about our Apple and Jamf offering: https://www.compnow.com.au/blog/jnuc22/ #JNUC2022 #Jamf #MacManagement

Categories
MacAdmin

MDM Managed Administrator Account

The tale of the macOS MDM Managed Local Administrator Account vs Jamf Management Account

Over the years as Jamf Pro and macOS have evolved, from pre-MDM framework, including the Casper Suite days, to the more recent evolutions of FileVault and SecureToken, Apple is investing more and more into “non-agent” frameworks to build on the Success of an MDM first approach in iOS.

Jamf Pro has been a fantastic tool for running policy and agent/binary based to fill in the gaps for where MDM framework initially didn’t existing, and then subsequent in its short comings.

The next low hanging fruit in both Apple and Jamf Pro’s evolution, around local macOS account management, is the macOS local administrator account.

Apple have recently clearly defined the future role of the “managed administrator account” that the MDM framework can remotely manage:

https://support.apple.com/en-au/guide/mdm/mdmca092ad96/web

Jamf Pro currently has a partial implementation of the “managed administrator account” as part of macOS PreStage Enrollment, however there currently is no ongoing “stateful” management of the account.

Jamf Pro does currently have a process of managing the password of Jamf Pro Management Account found in User-Initiated Enrolment using the Jamf Pro binary via policies.

A recent release of Jamf Pro better separated the MDM created PreStage enrolment account and the Jamf Management Account, however, the Jamf Management Account framework is largely one of Technical Debt in the Jamf Pro Framework.

2 Possible pathways forward:

  1. Migrate the Jamf Pro Management account out of policy/binary based management and assume the role of Apple’s “managed administrator account”. Some of the related Jamf Admin functions will need to be deprecated and some replaced by modern MDM features such as MDM enabled Apple Remote Desktop management
  2. Build out the MDM commands/framework for ongoing management of Apple’s MDM “managed administrator account” and mark the Jamf Management Account as deprecated. This would also involve replacing the Jamf Management account under UIE with the MDM “managed administrator account” for consistency across “Device Enrolment” and “Automated Device Enrolment” intended for corporately owned devices. User enrolment channel being developed by Apple will not have any management account in scope.

Which ever pathway is chosen, the messaging to Jamf Pro administrators in the community will be to move the primary corporate admin account account on corporately owned shared and one to one macOS devices to the MDM MDM “managed administrator account” and have a place on the Jamf Inventory Record to manage the password of the account as part of MDM commands and/or inventory data.

Similar to the concept of FileVault PRK and IRKs, I envision Jamf Administrators having the ability choose a common password across all devices, configured in one place, opted in as a default option on all macOS devices, with alternate options for individually specified and individually auto generated (ie LAPS concept) passwords on each computer inventory record. Auto generated, unique per machine, as found as an option with the Jamf Management account currently, should be a global option for the MDM “managed administrator account”.

The direction from Apple is clear and the technical debt of the Jamf Management account is confusing for many Jamf Administrators.

Here is a Feature Request I created before I turned it into a blog post (upvote away!):

https://www.jamf.com/jamf-nation/feature-requests/9590/macos-mdm-managed-local-administrator-account-vs-jamf-management-account

Here is a MacAdmins Community related discussion on the topic as well (non-Jamf specific):

https://twitter.com/wikiwalk/status/1275622118324162561

Categories
MacAdmin

VPP Redemption Codes & Apple School Manager

Another interesting discussion today on the MacAdmin’s slack revealed a workflow gap created for some schools when Apple deprecated Volume Purchasing (VPP) Redemption Codes.

Essentially, a really horrible process could be used to buy a bunch of licenses for an app, in the form of codes, and give them to end users to redeem.

It was superseded some time ago by Managed Distribution, championed by MDMs, to initially assign licenses to devices, “activated” against their Apple ID. This was later improved again by assigning directing to a device (no Apple ID required).

This evolution saw the decline of ye old redemption codes to the point that Apple chose to sunset it (for EDU only??) and focus on managed distribution. This has left a gap in workflow for some schools.

Some schools were using codes as a lightweight touch to tackle the ever popular adoption of bring your own device (BYOD), gifting apps to students to use on their personal devices (assume wrapping up in school fees). No need to enrol a BYO device into MDM.

With that option now gone, solidified by Apple forcing migration for the legacy volume purchasing portal to Apple School Manager in December 2019, schools are trying to figure out how to replace this workflow. Mass purchase of iTunes cards is being floated.

One option, which does involve MDM, is the new user enrolment MDM channel. I won’t go into detail here, but effectively iOS 13 and macOS 10.15 devices can enrol into your MDM using a managed Apple ID (from ASM) and get a quarantined slice of your device storage to install organisation content (if your MDM supports it). The MDM can’t even see your device serial number… making its new set of limitations a much more comfortable pill to swallow than “letting you install an app gives you access to erase my entire personal device” level of control.

The other option (which will be the most attractive to the redemption code loving crowd) is Apple Configurator 2.

This article points out a nice solution for “If you want to use managed distribution, but don’t have an MDM solution”:

https://support.apple.com/en-au/HT202995

Given you only need initial access to the device and then can revoke later as needed, this might be a nice solution.

To Add: https://support.apple.com/en-gb/guide/apple-configurator-2/cadbf9c811/mac

To Revoke: https://support.apple.com/en-gb/guide/apple-configurator-2/cadeaa4649f2/mac

Let’s see if this approach gets any traction with the BYOD wrangling EDU community.

Categories
MacAdmin

AddTrust Root CA Expiry and macOS

Update 2020-06-11: prior to May 30th I observed another symptom of this cert expiry that I didn’t comment on originally in this post. A client I recently worked with, who uses Aruba Clearpass to manage BYOD device onboarding to their managed WiFi SSID, were seeing this message across their user base.

The conclusion was the user context profiles installed manually by the user via the user driven onboarding process (which included the AddTrust root CA) were causing macOS to warn them around 30 days out and periodically after. Device level profiles including the same CA installed via Jamf Pro using SCEP did not alert the user. The CA was actually not in the current/relevant trust cain, but because it was managed via the profile, it alerted the user.

Update 2020-06-10: MacMule wrote an interesting post on the effects of this expiry on popular MacAdmins tool AutoPkg: https://macmule.com/2020/06/02/autopkg-curl-exit-status-60/

There have also been reports of on premise Jamf Pro environments having fallout by way of failed binary installs. When enrolling via either Automated Device Enrolment (DEP) or User-Initiated Enrolment the InstallApplication phase would likely deliver the initial package/commands to download the binary but the subsequent curl of binary components and binary enrolment will fail. This is most obviously identified by the MDM Profile and PPPC profile being installed but no other profiles and nothing under /usr/local/jamf (missing). The binary (& Self Service) were not present, causing the machine to fail enrolment and be present in the Jamf Pro web admin but marked as “unmanaged”.

Update 2020-06-03: there is a great write up with more technical detail at https://calnetweb.berkeley.edu/calnet-technologists/incommon-sectigo-certificate-service/addtrust-external-root-expiration-may-2020

This wonderful piece of info took off in Twitter and MacAdmins Slack today:

https://twitter.com/sleevi_/status/1266647545675210753?s=20

TLDR; The AddTrust root CA expired May 30 2020 and now OpenSSL libraries used in tools like `curl` are struggling to recognise intermediate certs that are cross-signed to get around expiring root issues

Your Mac will trust the cert in Safari, but curl (used to download things in scripts) may not for example.

Why this is a problem for macOS: https://mobile.twitter.com/sleevi_/status/1266781570108723208

It appears that macOS transition to LibreSSL as early as macOS 10.13 for some components but the bits left effect this bug today. `nscurl` is Apple’s variant and the basis of other tools in macOS does not seem to be affected.

Bottom line: check your scripts…. it appears I may have some work to do.

Categories
MacAdmin

Managing Apple Software Updates in macOS

For those managing Apple Software Updates with Munki, Jamf Policies or other scripted/binary/agent methods powered by the `SoftwareUpdate` binary, take note of this post from the team behind Munki:

https://github.com/munki/munki/wiki/manual-apple-updates-in-munki-5

TLDR: we should be using MDM commands on DEP enrolled machines and/or managing Apple software updates exclusively with configuration profiles, directing all updates through the native macOS process in System Preferences.

Here is a Jamf Perspective, but missing the above detail on software update binary issues:

https://docs.jamf.com/best-practice-workflows/jamf-pro/managing-macos-updates/Introduction.html

The reality is that any management system/script/etc that has historically used the softwareupdate binary is not a viable way of running updates moving forward (unless Apple change their trajectory, which I highly doubt).

If your org does not currently manage how System Preferences handles Apple Updates, start. If you are not currently using Automated Device Enrollment (DEP) or an MDM that can push Software Update commands over MDM framework, change.

Start managing your macOS System updates like iOS 😉

Update 2020-06-28: here is a more specific explanation of why the command line installer tools have become unreliable and a possible workflow for Jamf Pro (or similar tools) when you want to manage updates at specific intervals:

https://babodee.wordpress.com/2019/07/23/handling-macos-software-updates-with-jamf-pro/

Categories
MacAdmin

macOS Setup Assistant Preferences (Skip Screens)

As I have recently discussed with some of my colleagues, there have been some inconsistencies over the years managing the Setup Assistant experience on macOS when new users login. Lets talk about 2 ways many know to manage this:

Here are the 3 issues I have encountered:

  • All items being skipped for a new user login on a pre-configured computer, but the setting up your Mac screen still being displayed
  • All except a couple items being skipped for a new user login on a pre-configured computer, even though we “ticked all the skip options in the DEP Enrolment Profile”
  • Deploying the community sourced custom profiles above but only some of them work/don’t skip everything

Given I have been victim of all these scenarios I thought it was time to take a closer look.

First, to Apple’s documentation: https://developer.apple.com/documentation/devicemanagement/setupassistant

Stand out observations:

  • There are 8 preferences keys for skipping items
  • The profiles above only refer to 6/8, with one key per profile
  • Documentation and the community profiles both refer to the PayloadType needing to be com.apple.SetupAssistant.managed
  • Custom Settings Profiles from Jamf use com.apple.ManagedClient.preferences for the PayloadType when managing app preferences like those of Microsoft Office

So from this hit list of info, we see that our community profiles may be letting us down, possibly adding to one of our other issues we see as well.

Digging a bit further, I uploaded SkipiCloudSetup.mobileconfig to my test Jamf Pro and then re-downloaded and unsigned a copy of it to see if what I uploaded matched which I got back, it did’t. The important bits:

  • PayloadType was still com.apple.SetupAssistant.managed
  • SkipCloudSetup preference key was still there, but instead of being TRUE, it was FALSE
  • SkipSiriSetup preference key magically joined it in the profile and was also FALSE

Looks like best practice for making sure Jamf doesn’t mess with your custom config profile by Signing/Encrypting it before upload still applies: https://www.jamf.com/jamf-nation/articles/648/deploying-custom-configuration-profiles-using-jamf-pro

This will undoubtedly be causing a bunch of issues and confusing things to no end, summary on custom profiles:

  • Custom profiles we often refer to don’t include all the keys
  • When they are uploaded to Jamf as is, unencrypted, they may not end up on the computer the way they started and therefore not work

So where does this leave us for the other 2 issues? A few simple thoughts:

  • macOS changes as point releases come, so the behaviour of the Skip preference keys change, and new keys have been added over the last few major releases
  • If the MDM deploys DEP Enrolment Profiles with the Skip items included for High Sierra, and then machine over its life is upgraded to Catalina, that profile is never re-installed to include the latest Skip items, even if the MDM has added them into the product. ts a one off set of settings that doesn’t change until the device is wiped or intentionally un-enrolled and re-enrolled
  • In the same thought, the checkboxes we see in a Jamf PreStage do not mirror the preference keys in Apple’s documentation so likely these have changed and evolved over time and the behaviour of what is linked to what has changed. Jamf may be deployed deprecated Skip keys for some items, for example, satisfying all of the requirements but not quite 100% causing the “setting up your mac” screen to display as it tidies up the loose ends, especially if the user hasn’t logged in since an OS update

With all of this considered, it looks like deploying a set of managed  setup assistant preferences to all machines ON TOP of the DEP Profile is desired to you want to be 100% sure the machine experience is what you intend.

Here is where we go off documentation…

As I said before, the PayloadType for these preferences, according to Apple, needs to be different to normal preference management we perform for other apps.

Testing on macOS Catalina 10.15.4 and 10.15.5b, here is what I have found:

Create a plist with all of the preference keys, upload it into a NEW Jamf Pro file into the Custom App Settings payload, and it works!

defaults write ~/Downloads/com.apple.SetupAssistant.managed.plist SkipAppearance -bool truedefaults write ~/Downloads/com.apple.SetupAssistant.managed.plist SkipCloudSetup -bool true

defaults write ~/Downloads/com.apple.SetupAssistant.managed.plist SkipiCloudStorageSetup -bool true

defaults write ~/Downloads/com.apple.SetupAssistant.managed.plist SkipPrivacySetup -bool true

defaults write ~/Downloads/com.apple.SetupAssistant.managed.plist SkipSiriSetup -bool true

defaults write ~/Downloads/com.apple.SetupAssistant.managed.plist SkipTrueTone -bool true

defaults write ~/Downloads/com.apple.SetupAssistant.managed.plist SkipScreenTime -bool true

defaults write ~/Downloads/com.apple.SetupAssistant.managed.plist SkipTouchIDSetup -bool true

plutil -convert xml1 ~/Downloads/com.apple.SetupAssistant.managed.plist

As long as the Preference Domain is set as com.apple.SetupAssistant.managed and the uploaded property list file displays as such, were good

{SkipScreenTime=true, SkipTouchIDSetup=true, SkipAppearance=true, SkipPrivacySetup=true, SkipCloudSetup=true, SkipSiriSetup=true, SkipTrueTone=true, SkipiCloudStorageSetup=true}
Jamf Pro Application & Custom Settings Profile Payload

What happens in the background, is we get a profile with a PayloadType of com.apple.ManagedClient.preferences with sub items of:

  • Key: com.apple.SetupAssistant.managed
  • Dictionary -> Key: Forced
  • Dictionary -> Key: mcx_preference_settings
  • All sub preference keys we expect below that

As we are now delivering this as a FORCED MCX Preference for the Machine and therefore ALL existing AND new users, rather than a “set once” config per machine that MIGHT deliver to all users, this should give a more consistent experience in theory.

Its contrary to Apple’s explicit documentation, but in line with their overall ManagedPreferences framework: https://developer.apple.com/documentation/devicemanagement/managedpreferences

As I said, I have tested this flow and it works as expected with new users created after the profile was installed from a UIE/UAMDM Jamf configuration on 10.15.4/5 as well as a DEP configuration on 10.15.4/5.

Testing today on 10.15.5, having ONLY the DEP Profile with everything skipped resulted in the Setting Up Your Mac screen for a new user login. Applying my profile created with the attached plist from my example commands above, dismissed the Setup screen entirely on first login.

Conclusion

Use your DEP profile to set your desired 1st user experience for Setup Screens and use the Custom Profile (type com.apple.ManagedClient.preferences via Plist upload in the case of Jamf) to manage it from a ManagedPreferences engine/framework for all subsequent users that login to the Mac, keeping in mind the timing of delivery MAY override the 1st user experience settings in the DEP Profile (more testing to be seen on that).

Hope this is as helpful for other as it was for me.

Jamf Pro Plist: https://gist.github.com/aarondavidpolley/9e41928c64203c6cd65ba0a02a37b77b

//Aaron

Categories
General News Inspirational MacAdmin

The Potential Good of COVID-19

If the age of Coronavirus lasts for an extended period of time it will force a lot of workplaces that “use computers for primary tasks” to turn all of their processes into remote achievable ones.

In other words, those with difficulty attending a workplace (due to a physical disability, weak immune system, geographic barrier, etc) will have access to more jobs and new jobs than was the case before #COVID19. It’s only a theory at this point but it’s very possible.

If a company in Sydney (read large city) removes the need for its employees to be in its Sydney office in order to do their daily tasks, someone in Byron Bay (read regional town) can be a full time employee working from home. It’s a very interesting concept.

About 85% of my work since I moved to a regional area late last year is for clients interstate that I physically don’t visit. Coronavirus has reinforced my ability to work 24/7 from home rather than the office 1hr away and it be more accepted (when I actually could have done it already).

Now that it’s expected (as we’ve all been told to work from home at my employer), it’s even better as all my internal meetings have moved online and my colleagues are more engaged to work 100% remotely, not just me. It’s awesome.

I think the outcome of this unintentional social experiment is actually going to be very positive. People will realise all of the tasks that can actually be done and are more efficient without physical access/contact with other people, places, & spaces.

People will also come to better value and prioritise the things that are best done with physical access/face to face. They will get better at doing remote tasks well.

We will be more intentional and effective both in person and remote.

This is my hope for 2020 and COVID-19.

Stay safe

Categories
MacAdmin

VMware AirWatch Munki Implementation Teardown

Hi All,

As some of you may know, I am a big fan of ye old macOS management tool Munki. Though NOT an MDM, it is a powerful tool for managing macOS device apps and preferences and loved by the MacAdmins community.  Please read through on the links above if you want to know more on those…

Anyway, for those who know what all of this wonderful stuff is and are curious on how AirWatch is using the beloved tool, read below:

 

AirWatch Munki Implementation

Core Folder:

/Library/Application Support/AirWatch/Data/Munki/

Binary:

/Library/Application Support/AirWatch/Data/Munki/bin/managedsoftwareupdate

Some standard install paths exist but not used; probably created by the binary on its first run.

/Library/Managed\ Installs/Cache
/Library/Managed\ Installs/catalogs
/Library/Managed\ Installs/manifests

Contents of core folder /Library/Application Support/AirWatch/Data/Munki/:

  • Managed Installs
  • MunkiCache
  • Munki_Repo
  • bin

Main preference file:

defaults read /Library/Preferences/AirWatchManagedInstalls.plist 
{
 AppleSoftwareUpdatesOnly = 0;
 ClientIdentifier = "device_manifest.plist";
 FollowHTTPRedirects = none;
 InstallAppleSoftwareUpdates = 0;
 LastCheckDate = "2018-04-25 08:28:59 +0000";
 LastCheckResult = 0;
 LogFile = "/Library/Application Support/AirWatch/Data/Munki/Managed Installs/Logs/ManagedSoftwareUpdate.log";
 LogToSyslog = 0;
 LoggingLevel = 1;
 ManagedInstallDir = "/Library/Application Support/AirWatch/Data/Munki/Managed Installs";
 OldestUpdateDays = 0;
 PendingUpdateCount = 0;
 SoftwareRepoURL = "file:///Library/Application%20Support/AirWatch/Data/Munki/Munki_Repo/";
 UseClientCertificate = 0;
}

Compared to normal preference file location:

/Library/Preferences/ManagedInstalls.plist

Curiously, this is not a standard function to change which preference plist file it reads.

The “Munki_Repo” in the plist file above is a local folder which the binary reads as the Munki Repository (different to a tranditonal install where that would be pointing to a remote server)

The following traditional Munki Repo folders exist:

  • catalogs
  • icons
  • manifests

Traditional folders not present:

  • pkgs
  • pkgsinfo

A non traditional folder exists in the repo:

  • MunkiData

MunkiData contains a munki_data.plist which appears to be their way of knowing whats installed by them (AirWatch) and therefore knowing what to remove or not when a device un-enrolls from management. File contents:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<array>
 <dict>
 <key>ComputedBundleID</key>
 <string>com.vmw.macos.Chrome</string>
 <key>ComputedBundleVersion</key>
 <string>66.0.3359</string>
 <key>ManagedTime</key>
 <date>2018-04-25T09:08:16Z</date>
 <key>RemoveOnUnenroll</key>
 <true/>
 <key>munki_version</key>
 <string>3.0.0.3335</string>
 <key>name</key>
 <string>Chrome</string>
 </dict>
</array>
</plist>

Here are the contents of my example manifest plist in the Munki_Repo/manifests folder:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
 <key>catalogs</key>
 <array>
 <string>device_catalog.plist</string>
 </array>
 <key>managed_installs</key>
 <array>
 <string>Chrome</string>
 </array>
</dict>
</plist>

And the example catalog file, which includes all pkginfo :

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<array>
 <dict>
 <key>PackageCompleteURL</key>
 <string>https://localhost:7443/Application/GoogleChrome-66.0.3359.117.dmg?url=https://cdnau02.awmdm.com/cn500.airwatchportals.com/18659/Apps/1329a943-833f-4ebf-b36e-0baeb9e58d83.dmg?token=st=1524646902~exp=1524733602~acl=/*~hmac=7fb9d93e5cae26b64a3ae09553f1314fc5971a3277445b57575e366a1844149e&amp;size=68680725&amp;bundleid=com.vmw.macos.Chrome</string>
 <key>RestartAction</key>
 <string>None</string>
 <key>autoremove</key>
 <false/>
 <key>catalogs</key>
 <array>
 <string>device_catalog.plist</string>
 </array>
 <key>category</key>
 <string>Software</string>
 <key>description</key>
 <string></string>
 <key>developer</key>
 <string></string>
 <key>display_name</key>
 <string>GoogleChrome-66.0.3359.117</string>
 <key>installer_item_hash</key>
 <string>8d050591d8bd465dcae2d60a8e699bce037d0ce51f5da4349eed78b626e9ce47</string>
 <key>installer_item_location</key>
 <string>GoogleChrome-66.0.3359.117.dmg</string>
 <key>installer_item_size</key>
 <string>67071</string>
 <key>installer_type</key>
 <string>copy_from_dmg</string>
 <key>installs</key>
 <array>
 <dict>
 <key>CFBundleIdentifier</key>
 <string>com.google.Chrome</string>
 <key>CFBundleName</key>
 <string>Chrome</string>
 <key>CFBundleShortVersionString</key>
 <string>66.0.3359.117</string>
 <key>CFBundleVersion</key>
 <string>3359.117</string>
 <key>minosversion</key>
 <string>10.9.0</string>
 <key>path</key>
 <string>/Applications/Google Chrome.app</string>
 <key>type</key>
 <string>application</string>
 <key>version_comparison_key</key>
 <string>CFBundleShortVersionString</string>
 </dict>
 </array>
 <key>items_to_copy</key>
 <array>
 <dict>
 <key>destination_path</key>
 <string>/Applications</string>
 <key>source_item</key>
 <string>Google Chrome.app</string>
 </dict>
 </array>
 <key>minimum_os_version</key>
 <string>10.9.0</string>
 <key>name</key>
 <string>Chrome</string>
 <key>postinstall_script</key>
 <string>#!/bin/bash

open /Applications/Google\ Chrome.app

exit 0</string>
 <key>unattended_install</key>
 <false/>
 <key>unattended_uninstall</key>
 <false/>
 <key>uninstall_method</key>
 <string>remove_copied_items</string>
 <key>uninstallable</key>
 <true/>
 <key>version</key>
 <string>66.0.3359.117</string>
 </dict>
</array>
</plist>

The thing that stands out the most above is the PackageCompleteURL key used. Basically, the normal behaviour for items is to look in the Munki_Repo/pkgs folder for the asset but since the repo is local, they redirect to their storage for the actual package download. They do it via some local proxying method which is quite interesting…

In my example above I made the item on demand (rather than auto installed) and set a post install script to launch Chrome after it was installed (so I would know when it happened).

In a native munki world, you would be using the Managed Software Center GUI app to choose items that are “option installs” to install them on demand. In the AirWatch world, the back end system is making everything a managed install when it hits Munki, just holding it back until the user initiates it on an AirWatch portal as we’ll see shortly.

Its also worth noting that logs and other items are located in the “Managed Installs” folder as normal, except in the “/Library/Application Support/AirWatch/Data/Munki/Managed Installs/“ location rather than “/Library/Managed Installs/“

 

Walking Through Install Process

I used the “MyOrg Apps” Web shortcut the AirWatch Agent placed in my dock after it was installed and I was taken to a portal where I could browse or search for Apps that were assigned to me. On the front page was Chrome, so pressed to install and confirmed.

The AirWatch agent then started to do its work (shown. by the menu bar icon blinking) and after a minute or so Chrome launched as per my post install script.

My web based self service app portal now shows Chrome as “installed”

 

Comments On Preparation/Upload Process

The other interesting thing to note in my example is when I uploaded a DMG of the Google Chrome app into the AirWatch portal and assigned it, it asked me as part of the upload to download and install their “VMWare AirWatch Admin Assistant” on to my admin Mac to generate metadata.

The app basically asked for the DMG and less than a minute later spat out a folder in ~/Documents/VMware AirWatch Admin Assistant/ with an icon, DMG with modified name, and a plist containing the metadata the admin portal was after.

I would say in future it would be wise to run the assistant first and use the DMG it creates as I assume it makes sure the app in question is in the root level of the DMG as Munki prefers (different to the full path to place method Jamf uses for DMGs for example)

 

Final Thoughts

Overall, this is a simple but effective implementation of Munki leveraging the binary’s smarts but adding some integration with AirWatch systems to leverage the entire toolkit. It will be interesting to see how this aids AirWatch’s adoption for macOS management in the enterprise over the coming months/years.

Categories
MacAdmin

JamfWATCH

Hi All.

After being at a recent Jamf course I was inspired to create a new project called JamfWATCH.  As per GitHub:

Jamf Pro WATCH Dog: Monitor and self heal Jamf Pro enrolment if framework is removed from a client computer

Basically, at both reactive and recurring intervals this tool checks if it can communicate with the Jamf Pro Server its supposed to be managed by and if senses an issue tried to repair itself. Great for scenarios where end users may have admin rights and decide to do silly things like remove management from their computer.

Check it out, test, and provide feedback!

https://github.com/aarondavidpolley/JamfWATCH