Multiple services modifying attributes in Active Directory
In our Active Directory migration projects and IDM implementations we often come to situations where we need to have a look at the metdata of Active Directory attributes. When different synchronization services can modify attributes and local IT administration is ongoing, it is helpful to see very quickly which attribute was changed when and on which domain controller.
REPADMIN and PS-REPADMIN
Assumption made, that running synchronization services like DSA from Dell Migration Manager for Active Directory and maybe Forefront Identity Manager use different Domain Controllers, the object and attribute metadata can help you to sort out what was the latest change and by which tool in which domain.
When we were forced to use the native CL tool REPADMIN over and over to get evidence of what and when changed group membership, we decided to create the GUI based PS-REPADMIN utility.
The task of getting the Active Directory metadata at a glance for one identity in one or two domains – including the actual attribute values – is now easier to handle and see.
In comparison mode, you can see the metdata of a cross-forest synchronized object side-by-side and find out lack of synchronization streams.
The utility is built with Sapien Powershell Studio and fully based on Powershell and .NET. It requires Powershell 3.0 and the Active Directory Module to be present on the executing host. Active Directory Management Gateway Service needs to be present on Windows 2003 and Windows 2008 domain controllers, while no special requirments are necessary for Windows 2008R2 and later.
The utility was designed and tested for user and group objects.
The executing account needs to have read permissions in all domains on the objects you want to query for the metadata.
We used Quest Migration Manager 8.10 recently in a project at a customer for a combined Active Directory and Exchange migration. Overall target was to integrate a Windows 2003 domain cross forest and cross org into the central AD Forest with several child domains. Since from mail perspective our migration source was Exchange 2007 and our migration target Exchange 2013, we decided to use the Native Move Job option along with the Migration Manager for Exchange Agent (MAgE) services.
The customer environment look like the following:
Source Domain in Single Domain Forest with Domain Controllers on Windows 2003 and Exchange 2007 as mail system.
Target Domain was one of several child domains in the central Forest. All domain controllers running Windows 2012 R2 and mail system was Exchange 2013 SP1.
All Exchange 2013 servers had been deployed to root domain which also kept all important system and admin accounts.
To limit complexity in the setup of Quest Migration Manager 8.10, we decided to use a single administrative account from target Forest’s root domain and granted all necessary permissions in the domains to run both, Active Directory and Exchange migration. Only for access to source Exchange 2007 when running the move request, we used an account from source domain with Org Admin permissions.
Installation of Migration Manager 8.10. on a member server in target domain (best practice recommendation) including all cumulative hotfixes went smoothly. After successful Directory Synchronization, we connected to the Exchange source and target Organization and finally deployed 2 Instances of the MAgE agent for native mailbox move jobs on our agent host and console server. Note: For agent hosts Windows 2012 R2 is currently (May 2014) not supported. You have to stay with Windows 2008 R2 here.
However, after starting the agent services running with our administrative account , we recognized, that we could not open the log file of the agent in the Log Panel inside the Migration Manager for Exchange GUI. We searched for the log file and found it in “c:\progamdata\quest software\Migration Agent for Exchange\NativeMove directory”.
The log file showed that the agent was not starting to process the migration collection due to missing settings and then went to sleep. The lines of error:
Waiting for agent settings: Not found: (&(objectClass=serviceConnectionPoint) …..
Agent is not ready to start. Agent going to sleep at 1 minute.
repeated over and over.
Obviously the agent tried to execute an LDAP query to find a connection point in Active Directory. Note: Currently QMM 8.10 uses 3 different systems to store configuration data: An ADLDS server, a SQL Server Instance and the Active Directory (ADDS).
Service Connection Point (SCP):
We ran the query which was shown in the log file against the target domain and we could find the Service Connection Point (SCP) immediately in the System container of the domain naming context.
The Service Connection Point consists primarily of the keywords array attribute and the serviceBindingInformation attribute. The QMM MAgE looks for the serviceBindingInformation attribute to get its SQL connection properties. In SQL it will finally find all information to process the collection.
We do not know why Developers at Dell Software made this process so complex. However, in our setup the agent could not find the Service Connection Point, because the agent was looking in the domain, where its service account was located and this was the root domain of the forest while the agent host had installed the SCP during installation in the child domain where the computer account was member of.
Switching the agent host and agent service account to an account from child domain would have been a solution, but was not in compliance with customer policy to host all system accounts in root domain.
Moving agent host and console to root domain would not have meet best practices and would have interfered running directory synchronization.
So we ended up in giving the agent just what it requested:
We manually created a Service Connection Point in the root domain and copied all serviceBindingInformation values over.
The agent started immediately and worked without errors.
For future design we can only recommend to store Service Connection Point in the Configuration Partition as Exchange and lots of other software. Using the domain naming context will always lead to problems in a big Enterprise environment with Active Directory consisting of multiple domains in a forest.
However, the mixture of Powershell Versions we find at customer Environments will get wider and wider. Same is valid for modules like ActiveDirectory or SnapIns for Exchange. One will need to start with a lot of checks in the beginning of the code when a script is planned to be used universally.
In our large scale Active Directory Cross Forest migration project, we now have migrated already 40.000 user accounts globally. Our self made scripting routine to migrate/write sidHistory into the target accounts turned out to be a robust, reliable part of the process and I feel safe now to share some experiences. We are running it on multiple migration servers around the globe as scheduled task – which you can easily call a “service” as it is running every 5 minutes.
I will write about the whole mechanism of how we automated our large scale Active Directory migration in another blog post, but will concentrate here to share our way of managing the sidHistory part.
As you know already from part 2 of this blog post, we were buidling our code on the examples that MSFT Jiri Formacek published here.
However, 2 main restrictions prevented us from using this code as is:
We wanted to make sure that we really used the Domain Controller with the PDC Emulator role from source domain. Our source environment has 100+ domain controllers and the PDC role is siwtched from one DC to another DC under certain conditions. Therefore to use a fixed name for the PDC role Domain Controller was not acceptable.
Our Active Directory account migration process was fully automated and it was the user who starts his/her migration not us. Therefore the requirement was given, that we only can run sidHistory migration (together with the account activation in target domain) as a continuous background service. Every session based approach would not have helped like we can find it in ADMT or Dell Migration Manager for Active Directory.
Prepopulating sidHistory on the previously created disabled accounts in target domain was not an option, since Exchange 2010 was giving errors for disabled users with sidHistory of source active users under certain circumstances.
1) This was not a big thing. A small function could do the trick.
2) This was not that easy (for us). Running our account migration script as usual – means as scheduled task with admin credentials – did not work for the sidHistory part in it since the credentials of the logged user account were not handed over to the SIDCloner routine.
All the code we could find on Jiri’s page asked for credential information interactively or would need explicit credentials in the script in another way.
Although we are packaging our Powershell Scripts into an .exe file by using Sapien Powershell Studio and could hide the password from simple file editing, putting user name and password into the script was not an acceptable way for us to go.
After testing back and forth, someone cam up with the idea of using the Windows credential manager to work around our deadlock situation.
The script would access the credential manager interface, get the credential information from there and would then pass them to the DsAddSidHistory function.
We created a function to retrieve credentials from Credential Manager store based on a very good script example to be found on Technet here.
While this seemed to be a clever way of achieving our target of having a scheduled user account activation script with sidHistory functionality, we ran into errors again. Retrieving credentials from Credential Manager by script obviously fails, when the script runs with exactly the credentials that you want to retrieve. This was true in our case, because the user account migration script was scheduled with that “big admin” account.
The solution finally was: The user account migration script was running as a scheduled task with full admin credentials. When it came to migrate (in our project setup: activate) a user account in the target domain, it did not (could not) write sidHistory, but created an input file with username and target DC (the DC closest to the site where the user was had logged in from – remember that the user triggers his/her migration in our project).
On the same migration server a second script was scheduled with a server-local admin account. This script consists of 3 parts. First part is to check if there are new input files. Second part is to retrieve the full admin credentials from Credential manager and passing them to second part. Third part is to migrate sidHistory which succeeds because you have put all parts together for the SIDCloner routine:
PDC Emulator DC for source you have found by query.
Target DC was in file (but you can take every writable you want if replication delay does not matter).
Explicit credentials you get from Credential Manager.
Nowhere in both scripts password information is saved in clear text.