Monday, 16 July 2018

.NET Framework 4.7.2 Breaks AAD Connect and Exchange

Hi There,

Time for a new post.

Microsoft made .Net Framework 4.7.2 available on Windows Update on 10 July 2018, just about a week ago. As an "Important / Recommended" update, it gets under the radar at many organizations where all "Important" updates are installed as default practice. .NET updates used to come as "Optional". This time, however, Microsoft deemed this update "Important" for whatever odd reason that escapes me.

Although Microsoft "strongly recommends" the installation of this update, reports have emerged that it doesn't play nicely with AAD Connect. and Exchange. Specifically, CPU utilization of the Microsoft.Identity.Health.AadSync.MonitoringAgent.Startup.exe process goes through the roof, grinding the server to a halt:

Secondly, Microsoft has not (yet) updated the Exchange server prerequisites to reflect support for .NET Framework 4.7.2 - see

Sure enough, the update bricks the Exchange OWA and ECP portals too. After you log on, you get a pristine, white browser window, devoid from anything:

I thought OK, let's rebuild some virtual directories. Well, for that I need EMS - as long as it works. It fell flat too:

In fact, looking at the IIS logs, it becomes clear that pretty much everything has gone south.

As recovery steps, first I removed .Net 4.7.2 as some sources indicate on the Internet. Unfortunately that didn't fix the AAD Connect high CPU problem - it returned after an hour or so. And it certainly didn't fix the Exchange problem.

As far as Exchange is concerned, I tried the following:

  • Removed .Net 4.7.2
  • Removed and reinstalled .Net 4.7.1
  • Installed Exchange 2013 CU21 - the server was a tad outdated, on CU13

No joy. The screenshots above were taken after the recovery attempt.

My recommendation to you, dear reader, is to block the installation of .Net 4.7.2 for the time being. It is NOT an "important" update, no matter how much Microsoft would like you to believe.

The update can be blocked with a Registry setting, as documented at KB4342394.

I am in for rebuilding the Exchange server bricked by Microsoft's (not so) "important" .Net update. Thank you Mr. Microsoft, yet another .Net blunder to add to the list.

Happy patching!


Microsoft has come to its senses and re-published .NET Framework 4.7.2 where it belongs, under "Optional" updates.

Friday, 20 April 2018

Changing the Garbage Collection Mode on MSExchangeMapiFrontEndPool

Hi There,

I stumbled on a script at, referenced by

Ran the script on an Exchange 2013 system, and among others, it threw an error for the MSExchangeMapiFrontEndAppPool using Workstation garbage collection mode instead of Server mode:

Not being a .Net developer, looked up the issue, and all sources that I could find indicated that the garbage collection should be configured for Server mode for improved performance. Just as the script said.

I was (and still am) gobsmacked as to why the Exchange team chose this mode while everything else is optimised to the extreme.

The script is great, kudos for Marc Nivens, except it doesn't provide any reference as to how to go about changing the garbage collection mode, which would be very helpful for folk like me who aren't developers.

So here is how to do it:

First, we need to find where garbage collection is configured. For that we need to identify the configuration file for the .Net application. There are a number of ways, two of which are shown below:

1. The DOS way:

%WINDIR%\System32\Inetsrv\appcmd list apppool "MSExchangeMapiFrontEndAppPool" /text:"CLRConfigFile"

2. The PowerShell way (in case you want to script it):

[System.Reflection.Assembly]::LoadWithPartialName("Microsoft.Web.Administration") | Out-Null
$serverManager = new-object Microsoft.Web.Administration.ServerManager
(($serverManager.ApplicationPools | ?{$_.Name -eq "MSExchangeMapiFrontEndAppPool"}).Attributes | ?{$_.Name -eq "CLRConfigFile"}).Value

Essentially the file is %ExchangeInstallPath%\bin\MSExchangeMapiFrontEndAppPool_CLRConfig.config

Open the file in an elevated Notepad and change the <gcServer enabled="false" /> line to <gcServer enabled="true" />. Here is the file with the default value:

After the change (and a server restart) I got this status:

A word of caution: When a new CU is installed then it overwrites the config file, thus reverting back the garbage collector to Workstation mode. Don't forget to update the file again after the CU is installed.


Happy garbage collecting!

Monday, 25 December 2017

Assign or Remove O365 Licenses

Hi Folks,

Here is a script for managing O365 licenses. Get it from the TechNet Gallery.

If you work with O365 then sooner or later you''ll be asked to assign some licences - or remove them to that matter - to more than just one user. Using the portal will take forever.

A PowerShell script is the natural option, but it may be daunting to get the hang of the licensing object structures and to get your head around building a DisabledOptions list of services of what NOT should be enabled as opposed to just say "this is enabled, that is not".

I've put together a script which will do it all: it will enable or disable a particular service within a plan.

The script implements an interactive, user-driven flow of actions, which resembles the way NTDSUTIL or DISKPART work.

Prerequisites: You'll need the Microsoft Azure Active Directory Module for Windows PowerShell.

How it works:

1. Create a text file with all the UPNs for all the users who will have a license assigned or removed. No header line, just a list of UPNs:

In this example we have 3x valid users, 1x non-existent user and 1x incorrectly formatted user - will see its significance later in this post.
2. Run the script:

Manage-O365Licenses.ps1 -InputFile "Your_Text_File_With_UPNs.txt" -Location CountryCode

where the country code is the ISO 3166 Alpha-2 two-letter code of the country that the users will be assigned to. For the full list of country codes see

For example, if your users' UPNs are listed in Users.txt and they are all in Australia, you would use the following command ...:

Manage-O365Licenses.ps1 -InputFile Users.txt -Location AU

... or simply launch the script with no parameters at all and it will prompt for the file and location:


NOTE: If your users are distributed across different countries then you'll have to run the script separately for each country. It doesn't allow for mixing locations in one run.

If you happen to be already connected to O365 (you ran Connect-MSOLService previously and authenticated successfully) then the script will move on to the next step. Otherwise it will prompt for credentials, then it will connect to O365.

3. Select what you want to do: Disable or Enable licenses:

Select 0 or 1 and press Enter, or X to exit.
Your selection will be confirmed in green (see next screenshot).

4. Select the plan. Same process: enter a number, or X to exit:

Some plans are exposed in PowerShell, however they are hidden in he GUI. I couldn't identify a reason. If it isn't in the GUI then probably it shouldn't be touched. These show up as "--- not available for selection in the Portal ---".

Also, the list only contains the plans that I have come across, and it is probably incomplete, or Microsoft will come up with new ones. If you come across such a plan then its cryptic name will be displayed with "(no friendly name found)" appended, similar to the following:

For the record, the screenshot is a simulation of what you may encounter. The SPE_3 plan is the equivalent of "Microsoft 365 E3", and I removed the definition for the sake of illustration. It is back in the version that's available for download in the TechNet Gallery.
5. Select the service. Same process: enter a number, or X to exit. For illustration, I typed a couple of invalid selections to demo how it is handled:

6. Then a last chance of change of heart: review your selections and select Y or N:

The script runs and displays the statistics:

What do the stats mean? Let's open the license.log file:

There is an entry for each user:

  • SUCCESS: The operation has completed successfully for the user.
  • NOCHANGE: The license is already at the desired state, no change needed.
  • FAIL: The user does not exist in the organization (e.g. or it failed format validation (e.g. gibberish is not in the correct format).
  • NOAVAILABLELICENSE: Ran out of licenses (not illustrated here).

WARNING: The script overwrites the log file! If you want to preserve the logs between different runs then move or rename the file before you run the script again.

The scripts will also catch some errors. In case an error is caught, a debug.log file is created. It is only created if a cmdlet generates one of those red errors.

Just like the license.log file, debug.log is wiped every time the script runs. However it is only created if a cmdlet generates an error. It looks like this ...:

... and its associated entry in the license.log file:

Some error handling features:

  • Validation of input file. Exits if it doesn't exist.
  • Validation of country code. Exits if it is invalid.
  • Graceful exit if credential input is cancelled.
  • Graceful exit if wrong credentials are provided.
  • Graceful handling of invalid selection input (out of range or invalid characters).
  • Option to exit at any point.
  • UPN format validation.
  • Existing user validation.
  • Licence count validation.
  • Catches some error conditions with error details in debug.log. No error, no log.
  • Detailed record of changes and other stats in license.log (not an error but good to have).

Licence Agreement

The scripts are licenced under the NMF (Not My Fault) agreement.
Edit, change and use the scripts at your own risk.
Give credit where due.

Download: Manage-O365Licenses.ps1

Enjoy :-)

Saturday, 11 November 2017

Exchange 2016 CU7 Bug Causes HCW to Fail

Hi There,

If you plan to deploy a hybrid Exchange server using Exchange 2016 CU7, then don't. It's broken. Use a pre-CU7 server instead.

A bug slipped into Exchange 2016 CU7 which prevents the HCW from completeing. The HCW fails to get past the domain ownership validation:

No matter how hard you try, you can't get past this screen.

Looking at the HCW log at C:\Users\<username>\AppData\Roaming\Microsoft\Exchange Hybrid Configuration, the following errors are logged:

*ERROR* 10300 [Client=UX, Thread=1] System.ArgumentOutOfRangeException: The UTC time represented when the offset is applied must be between year 0 and 10,000.
                                Parameter name: offset
                                   at System.DateTimeOffset.ValidateDate(DateTime dateTime, TimeSpan offset)
                                   at System.DateTimeOffset..ctor(DateTime dateTime)
                                   at Microsoft.Online.CSE.Hybrid.App.AppData.BuildSessionProperties(NotificationType type)


*ERROR* 10277 [Client=UX, Activity=Domain Ownership, Session=OnPremises, Cmdlet=Set-FederatedOrganizationIdentifier, Thread=14] FINISH Time=2598.6ms Results=PowerShell failed to invoke 'Set-FederatedOrganizationIdentifier': Object reference not set to an instance of an object. An unexpected error has occurred and a Watson dump is being generated: Object reference not set to an instance of an object.


*ERROR* 10224 [Client=UX, Page=DomainProof, Thread=14] Microsoft.Online.CSE.Hybrid.PowerShell.PowerShellInvokeException: PowerShell failed to invoke 'Set-FederatedOrganizationIdentifier': Object reference not set to an instance of an object. An unexpected error has occurred and a Watson dump is being generated: Object reference not set to an instance of an object. ---> System.Management.Automation.RemoteException: Object reference not set to an instance of an object.
                                   at System.Management.Automation.Runspaces.AsyncResult.EndInvoke()
                                   at System.Management.Automation.PowerShell.EndInvoke(IAsyncResult asyncResult)
                                   at Microsoft.Online.CSE.Hybrid.Provider.PowerShell.PowerShellProvider.PowerShellInstance.Invoke(String cmdlet, IReadOnlyDictionary`2 parameters, Int32 millisecondsTimeout)
                                   --- End of inner exception stack trace ---
                                   at Microsoft.Online.CSE.Hybrid.PowerShell.RemotePowershellSession.RunCommandInternal(Cmdlet cmdlet, SessionParameters parameters, Int32 millisecondsTimeout, PowerShellRetrySettings retrySettings, Boolean skipCmdletLogging)
                                   at Microsoft.Online.CSE.Hybrid.Session.PowerShellOnPremisesSession.SetFederatedOrganizationIdentifier(SmtpDomain accountNamespace, String delegationTrustLink, SmtpDomain defaultDomain)
                                   at Microsoft.Online.CSE.Hybrid.App.ViewModel.Pages.DomainProof.DomainInfo.AddFederatedDomain(IOnPremisesSession session, AppData appData)
                                   at System.Collections.Generic.List`1.ForEach(Action`1 action)
                                   at Microsoft.Online.CSE.Hybrid.App.ViewModel.Pages.DomainProof.VerifyActivity(IOnPremisesSession session, EnvironmentBase environment)

Issues with the HCW have also been reported by admins on Rhoderick Milen's blog at

Microsoft is aware of it, and a fix will be included with CU8. Some suggest using a pre-CU7 server to run the HCW on - I haven't tried, not (yet) an option in the affected environment.


  • Tried running the HCW on a CU6 server that I installed temporarily. Since the CU7 server was still selected as the hybrid server, the wizard FAILED.
  • Tried the workaround suggested by AloneInTheDarK. While some reported success, it didn't work for me - see Microsoft advised that it is not a recommended workaround, even if it would have worked.
  • Downgraded the CU7 server to CU6. SUCCESSFUL. It involved:
    • Installed a temporary CU6 server to host all the mailboxes specific to Exchange 2016 while the CU7 server was downgraded (Arbitration, AuditLog etc. - see and see This is needed so that the CU7 server can be uninstalled.
    • Uninstalled Exchange 2016 CU7 and deleted the binaries from the installation folder.
    • Installed Exchange 2016 CU6 on the same machine.
    • Configured the downgraded server (virtual directories, certificate etc.), and moved all system mailboxes back to it..
    • Uninstalled the temp Exchange 2016 server.
    • Ran the HCW wizard - SUCCESSFUL
It looks like Microsoft has a hotfix. At this time I don't know whether it will ever be published as CU8 is due in the near future, but if you are in dire need then open a case and request the hotfix. I didn't get to test because it came in one day late, after I went through the downgrade ordeal.

Take care,

Friday, 20 October 2017

Don't Use the 32-bit Version of PowerShell on Windows Server 2016!

You fire up your new Server 2016 and want to install the Exchange 2016 prerequisites. You launch PowerShell and discover that Install-WindowsFeature isn't found:

Install-WindowsFeature : The term 'Install-WindowsFeature' is not recognised...

You want to import the ServerManager module, only to be greeted by another error:

Import-Module : The specified module 'ServerManager' was not loaded because no valid module file was found in any module directory.

You do all sorts of black magic to make ServerManager load, perhaps you copy C:\Windows\System32\WindowsPowerShell\v1.0\Modules\ServerManager to C:\Users\<your_account>\Documents\WindowsPowerShell\Modules\ServerManager. You'll eventually get to load ServerManager, and you breathe easy - until you try to install the prerequisites. Then comes the shock:

Install-WindowsFeature : The target of the specified cmdlet cannot be a Windows client-based operating system.

WHAAAT? Heck, your're machine is Server 2016!

Well, my friend, did you notice the "(x86)" bit in your PowerShell window's caption?

Yes, you're running a 32-bit PowerShell on a 64-bit server.

I can't blame you, it's easy to just type PowerShell and hit Enter. The problem is that the default selection may bring up the link for the wrong architecture:

The fix? Launch the 64-bit version of PowerShell and you'll be laughing.


Wednesday, 23 August 2017

NUMA Optimization on HP Proliant Gen9 Servers


I've just come across Ingo Gegenwarth's post about NUMA optimization on HP ProLiant Gen9 servers that may boost a server's performance significantly.

Read his post at


O365 IDM Bug: DirSync Errors but No Objects in Error

Hi All,

First things first, a Big Red WARNING: The following content applies to the specific context encountered in one of my client's O365 tenant. If you think you should follow it, please be extra careful that you understand the whats and whys of this post. It worked in my environment for my situation. It may not work in yours, or you may have to take additional steps to avoid data loss. If not careful, you MAY LOOSE DATA.

I do NOT accept responsibility for any data loss. Proceed at your own risk.

I recently worked on an issue where the O365 admin portal reported DirSync errors...,

... but when I looked at the detailed list of objects in error, the list was blank:

While I don't know what exactly happened, here is what I think the customer has done, as it makes the most sense:
  • Some users left the company. Their external contacts kept sending e-mail, and these e-mails had to be delivered somewhere so that business wasn't affected.
  • The client's admin decided to delete their mailbox and AD object. Created a distribution list, assigned the departed user's e-mail address, and populated the distribution list with recipients assigned to handle incoming e-mails for the departed users. No idea what happened to the departed users' mailbox content, and it is totally irrelevant for the purpose of this blog.
  • AADConnect deleted the O365 objects by means of AD sync. Consequently O365 moved objects into the Dumpster.
  • AADConnect tried to synch the new distribution groups, however an email address conflict was detected between the objects in the dumpster. The sync succeeded but with errors. For more details see the Duplicate Attribute Resiliency feature described in the first bullet point in the References section at the end of this post.
From this moment on, no matter what was done, the DirSync error persisted.

To fix it, I had to:
  • Purge the old accounts from the O365 dumpster.
  • Delete the distribution groups that were in error from O365 by moving them out of the AADConnect sync scope and having it do an AD sync.
  • Recreate the distribution groups by moving them back into the AADConnect sync scope and having it do an AD sync.

Here comes the long story. You'll need the Windows Azure Active Directory Module for Windows PowerShell.

Step 1: Connect to the O365 tenant with a tenant admin account:


Step 2: List the objects in error:


The command gives us a couple of useful tips:
- a DisplayName
- the conflicting properties in the ProvisioningErrors field

I also searched recipients for any conflicting addresses - no joy.

I checked the O365 dumpster for any deleted accounts. Bingo! Some entries had property values also found in the objects in error.

Step 3: List the objects in the dumpster, along with their proxy addresses. Note that I used the -Wrap parameter in the FT (Format-Table)  command in case the list of proxy addresses is too long, to avoid truncation (type it all on one line).

Get-MsolUser -ReturnDeletedUsers | ft DisplayName,ProxyAddresses -Wrap

In my case the tenant deleted some user accounts for staff who moved on, and replaced them with distribution groups populated with other users who then got the deleted user's e-mails, assigning the departed user's address to the distribution group. Therefore the same e-mail address was on the new distribution group, as well as on the deleted object in the dumpster which wasn't yet purged.

Hint: Purging the dumpster and re-synching AD doesn't solve the problem. It looks like it's a bug in the O365 identity management system. We still need to purge the dumpster but some additional steps are needed, so read on.

Step 4: To remove objects from the dumpster, use the command (type it all on one line):

Get-MsolUser -ReturnDeletedUsers | Remove-MsolUser -RemoveFromRecycleBin -Force

The command has no output:

Once the dumpster is purged and a new AD sync is initiated (or wait for the next cycle), you would think it's all sorted. Unfortunately it's not so. Once you do a re-sync and issue the Get-MsolDirSyncProvisioningError command, you'll still see that the conflicting objects are still in error.

No, the sync process isn't (yet) that evolved. A few more steps to go to get it sorted.

Step 5: Delete the object that is reported by Get-MsolDirSyncProvisioningError as being in conflict:
  • In the on-premises AD, move the objects in conflict outside the scope of the AADConnect sync scope (you didn't set the scope to the entire domain, right?)
  • Initiate an AD sync or wait for it to happen on schedule.
  • Confirm that the objects have been deleted in the O365 tenant.
  • Check the dumpster again. If the objects were moved there, then purge them as detailed at point 4 above.

Step 6: Run the Get-MsolDirSyncProvisioningError command again. It should return nothing:

Step 7: Move the object back into the AADConnect sync scope in the local AD and initiate a sync.

Step 8: Confirm that the objects have been re-created in O365.

Step 9: Confirm that the Get-MsolDirSyncProvisioningError command returns no more entries, and that's also reflected in the portal:

That's it.

Please note that in my case I only had to deal with Distribution Groups. Deleting and recreating them in O365 did not involve any user data, contrary to mailbox objects. That would have complicated things. Please see my warning at the top of this post.


Happy error hunting!