Tag Archives: F5

Renewing SSL Certificates on ADFS 3.0 (Server 2012 R2)

I recently had to replace the public-facing service communication certificates on our primary ADFS deployment on Server 2012 R2. I followed a procedure that I thought had a reasonable chance of actually doing what I wanted it to:

  1. Obtained a new private key with signed certificate.
  2. Saved the file to a pfx, and imported it onto each node in the ADFS cluster
  3. Set permissions on the certificate according to documentation
  4. Used the ADFS MMC -> Certificates -> Set Service Communications Certificate.

Everything seemed to go okay, but after a bit we started to get some complaints that some of our users could not access the Office 365 Pro Plus software download page. This was a curiosity to me, because I could not reproduce the problem. A colleague later noticed a raft of SSL errors in the System event log on one of the ADFS nodes, and disabled it in the load balancer configuration.

When I finally got around to investigating, I noticed that the system log reported problems from source ‘HTTPEvent’, with details DeviceObject: DeviceHttpReqQueue, Endpoint: adfs.uvm.edu:443 (and also Endpoint: What gives?

I found a related TechNet Blog that shed some light on the subject:

According to this document, after setting the Service Communications Certificate in the MMC, you must run:


to fetch the certificate thumbprint of the Service Communications Cert. Take note of the certificate thumbprint, then run:

Set-ADFSSslCertificate -Thumbprint [yourThumbprint]

“Set-AdfsSslCertificate” will fix the HTTP.SYS bindings used by ADFS. Apparently the MMC does not set the bindings, which is pretty annoying because this leaves the service in a pretty darn broken state. The HTTP bindings are mentioned in this TechNet documentation:
BUT, the docs do not explicitly state that the Set-AdfsSslCertificate cmdlet needs to be run on all of the ADFS server nodes in your farm. This also is a key missing detail.

Good Documentation… you always take it for granted, until you don’t have it anymore.

Note above that I mentioned a binding problem with the address This was a carry-over from our initial deployment of ADFS 3. Back then, Microsoft did not provide a health check URL for ADFS, and the supplemental binding was needed to allow health monitor connections from our F5 load balancer without using SNI, which is required by ADFS 3.0, but not supported on the F5. These days (and if you have KB2975719 installed), you can instead monitor the following URL from your F5:


More details on this solution can be found here:

While it is nice having a proper health check, problems can arise when your ADFS server HTTP bindings go sour. It would seem that nothing is perfect.

Integrating SharePoint 2013 with ADFS and Shibboleth… The Motion Picture

Time again to attempt to implement that exciting technology, Federation Services (Web Single Sign On, SAML, WS-Federation, or whatever you want to call it) with SharePoint. Last time we tried this, SharePoint 2010 was a baby product, MS was just testing the waters with SAML 2.0 support in ADFS 2.0, and Shibboleth 2 was pretty new stuff here at UVM. The whole experience was unsatisfying. SharePoint STS configuration was full of arcane PowerShell commands, ADFS setup was complicated by poor farm setup documentation, and interop of Shibboleth 2 with ADFS 2 was not documented at all.  After wading though all of that mess, we ended up with user names being displayed as “i:05.t|adfsServiceName|userPrincipalName” (bleagh!), and with many applications that could not deal with Web SSO authenticators.  My general conclusion was that SharePoint really was not ready for federated login.

Four years later, things have changed a bit.  STS configuration still requires dense PowerShell commands, but at least it is better documented.  ADFS and Shibboleth interoperability also are excellently documented at this point.  Microsoft has most Office apps working with passive SAML authenticators, and has pledged to get the rest working this year.  While I would not judge the use of ADFS (with or without Shibboleth) to be the easy route to take to SharePoint 2013 deployment, it at least looks functional at this point.  So let’s kick the tires and see how it works…

Part 1: ADFS Setup

First, we need to setup ADFS.  We chose to deploy “ADFS 3” using Windows Server 2012 R2 as the OS platform.  ADFS 3 is required to support the new “workplace join” feature of Server 2012 R2.  Since we want to test this, we would need ADFS 3 anyway.  Unfortunately, ADFS on Server 2012 R2 is pretty virgin territory, and does not have the same troubleshooting resources available for it as do earlier releases.

Most of the configuration steps followed the TechNet documentation without variation.  We did find that we needed to modify the ACL on the service account used to run ADFS… we added the “Authenticated Users” principal to the ACL, and assigned the “Read all properties” right.

For us, the most complicating factor in ADFS 3 deployment is the replacement of IIS with the Windows Kernel-mode HTTP server “http.sys”.  When we started experiencing connectivity problems with various clients to ADFS, we had no experience with HTTP.sys to assist in troubleshooting.  Most articles on HTTP.sys relate to remote desktop services, and with Server 2003.  Our problems with HTTP.sys were rooted in an undocumented requirement for clients to submit SNI (Service Name Indicator) information in their TLS “CLIENT HELLO” sequence.  I had to open a support case with Microsoft to resolve this problem, and only afterword was I able to find any Internet discussions that reflected MS advice:

It seems that if http.sys is bound using the hostname:port format, then TLS will require SNI.  If the binding is instead specified using ipAddr:port, SNI will not be required.  To fix our problem, we just needed to add a second HTTPS certificate binding using an IP address.  In this case, we just used “”.  Here is the procedure:

  1. On each ADFS server and proxy, open an elevated command prompt
  2. run: netsh http show sslcert
  3. Record the certificate hash and application ID for the certificate used by ADFS
  4. run: netsh http add sslcert ipport= certhash= appid={}

Part 2: Configuring SharePoint to use ADFS

I started with Microsoft’s guide to configuring SAML authentication for SharePoint 2013 using ADFS:
This is a well written guide, and fairly easy to follow.  The only issue I take with the article is the recommendation to use the “emailAddress” claim as the “identifier claim” in SharePoint.  In many federated login scenarios, a foreign Idp may not want to release the email address attribute.  However, some variation of UPN likely will be released.  In the case of the InCommon federation, the “ePPN” value (eduPersonPrincipalName) generally is available to federation partners.  For this reason I chose to implement “UPN” as the “identifier claim” in the last command of phase 3 of the document:

$ap = New-SPTrustedIdentityTokenIssuer -Name  -Description  -realm $realm -ImportTrustCertificate $cert -ClaimsMappings $emailClaimMap,$upnClaimMap,$roleClaimMap,$sidClaimMap -SignInUrl $signInURL -IdentifierClaim $roleClaimmap.InputClaimType

Part 3: Migrating SharePoint Users From Windows to Claims

Since we are planning an upgrade, not a new deployment, we need migrate existing Windows account references in SharePoint to federated account references. To make this happen, we need to establish the federated account identity format. I simply log in to an open-access SharePoint site as an ADFS user, and record the account information. In this case, out accounts look like this:


and groups:


We then can use PowerShell to find all account entries in SharePoint, and use the “Move-SPUser” PowerShell cmdlet to convert them.  I am still working on a final migration script for the production cutover, and I will try to post it here when it is ready.

Of some concern is keeping AD group permissions functional.  It turns out that SharePoint will respect AD group permissions for ADFS principals, but there are a few requirements:

  1. The incoming login token needs to contain the claim type “http://schemas.microsoft.com/ws/2008/06/identity/claims/role“,
    and that this claim type needs to contain the SamAccountID (or CN) of the AD
    Group that was granted access to a site.  (In ADFS, this means that you need to release the AD LDAP Attribute “Token-Groups – Unqualified Names” as the outgoing claim type “Role”.
  2. When adding AD Groups as permissions in SharePoint, we need to use the “samAccountName” LDAP attribute as the identifier claim.  The LDAPCP (see Part 4) utility makes this easy as it will do this for us automatically when configured to search AD.

Requirement 1 could be a problem when using Shibboleth as the authentication provider.  Our Shibboleth deployment does not authenticate against AD, so a Shibboleth ticket will not contain AD LDAP “tokenGroup” data in the “role” claim.  I am working with the Shibboleth guys to see if there is any way to augment Shibboleth tokens with data pulled from AD.

Part 4:  The SharePoint PeoplePicker and ADFS

Experienced SharePoint users all know (and mostly love) the “people picker” that searches Active Directory to validate user and group names that are to be added to the access list for a SharePoint site.  One of the core problems with federation services is that they are authentication systems only.  ADFS and Shibboleth do not implement a directory service.  You cannot do a lookup on an ADFS principal that a user adds to a SharePoint site.  This is particularly irksome, since all of our ADFS users actually have matching accounts in Active Directory.

Fortunately, there is a solution… you can add a “Custom Claims Provider” into SharePoint which will augment incoming ADFS claims with matching user data pulled from Active Directory.  This provider also integrates with the PeoplePicker to allow querying of AD to validate Claims users that are being added to a SharePoint site.  A good write-up can be found here:

“But I don’t want to compile a SharePoint solution using Visual Studio,” I hear you (and me) whine.  No problem… there is a very good pre-build solution available from CodePlex:

Normally I do not like using third-part add-ons in SharePoint.  I will make an exception for LDAPCP because:

  1. It works.
  2. It saves me hours of Visual Studio work.
  3. It is a very popular project and thus likely to survive on CodePlex until it is no longer needed.
  4. If the project dies, we can implement our own Claims Provider using templates provided elsewhere with (hopefully) minimal fuss.

My only outstanding problem with LDAPCP is that it will not query principals in our Guest AD forest.  However, there are some suggestions from the developer along these lines:

To summarize, the developer recommends compiling our own version of LDAPCP from the provided source code.   We would use the method “SetLDAPConnections” found in the “LDAPCP_Custom.cs” source file to add an additional LDAP query source to the solution.  I will try this as time permits.

Part 5: Transforming Shibboleth tokens to ADFS

So far we have not strayed too far from well-trodden paths on the Internet.  Now we get to the fun part… configuring ADFS as a relying party to our Shibboleth Idp, then transforming the incoming Shibboleth SAML token into an ADFS token that can be consumed by SharePoint.

Microsoft published a rather useful guide on ADFS/Shibboleth/InCommon integration:
Using this guide we were able to set up ADFS as a relying party to our existing Shibboleth Idp with minimal fuss.  Since we already have an Idp, we skipped most of Step 1, and then jumped to Step 4 as we did not need to configure Shibboleth as an SP to ADFS.

I had the local “Shibboleth Guy” add our ADFS server to the relying parties configuration file on the Shib server, and release “uvm-common” attributes for this provider. This allows SharePoint/ADFS users to get their “eduPersonPrincipalName” (ePPN) released to ADFS/SharePoint from Shibboleth. However, SharePoint (and ADFS) do not natively understand this attribute, so we configure a “Claim Rule” on the “Claims Provider Trust” with Shibboleth.  The rule is an “Acceptance Transformation Rule” that we title “Transform ePPN to UPN”, and it has the following syntax:

c:[Type == "urn:oid:"]
=> issue(Type = "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/upn",
Issuer = c.Issuer, OriginalIssuer = c.OriginalIssuer, Value = c.Value,
ValueType = c.ValueType);

The “urn:oid:” bit is the SAML 2 identifier for the ePPN type.

After configuring a basic Relying Party trust with SharePoint 2013, I need to configure Claims Rules that will release Shibboleth User attributes/claims to SharePoint.  You could use a simple “passthough” rule for this.  However, I want incoming Shibboleth tokens that have a “@uvm.edu” UPN suffix to be treated as though they are Active Directory users.  To accomplish this, I need to do a claims transformation. In AD, the user UPN has the “@campus.ad.uvm.edu” suffix, so let’s transform the Shibboleth UPN using a Claim Rule on the SharePoint Relying Party Trust:

c:[Type == "urn:oid:", Value =~ "@uvm.edu$"]
=> issue(Type = "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/upn", Issuer = "AD AUTHORITY", OriginalIssuer = c.OriginalIssuer, Value = regexreplace(c.Value, "^(?[^@]+)@(.+)$", "${user}@campus.ad.uvm.edu"), ValueType = c.ValueType);

NOTE: Code munged by WordPress. Contact me if you need exact syntax!

This appears to work… the first RegEx “@uvm.edu$” should match an incoming UPN that ends with “@uvm.edu”. In the second set of regExps, we create a capture group for the user portion of the UPN (which is everything from the start of the value up to (but not including) the “@” character), and place the captured data into a variable called “user”. We then replace everything trailing the user portion with “@campus.ad.uvm.edu”.

However, as noted above, the incoming Shibboleth SAML token does not contain AD group data in the “Role” attribute, so users authenticating from Shibboleth cannot get access to sites where they have been granted access using AD groups.  Not good! Fortunately, there is a solution.

An only-hinted at, and certainly not well documented capability of the Claim Rule Language is the ability to create an issuable token with claims originating from more than one identity store. In our case, we need to supplement the incoming Shibboleth SAML token with “roles” claims obtained from Active Directory. I do this during the release of the ADFS token to the SharePoint relying party, using the following rule:

c:[Type == "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/upn", Value =~ "^.+@campus.ad.uvm.edu$"]
=> issue(store = "Active Directory", types = ("http://schemas.microsoft.com/ws/2008/06/identity/claims/role"), query = ";tokenGroups;CAMPUS{0}", param = regexreplace(c.Value, "^(?.+)@campus.ad.uvm.edu$", "${user}"));

NOTE: code munged by WordPress. Contact me if you need exact syntax!

The issuance part of this rule bears some discussion. We use the “query” string to collect tokenGroups (or recursive group memberships). The query is run against the Active Directory ADFS attribute store (hence store = “Active Directory”). The query must take the format:
In our case, there is no filter specified. According to TechNet, the default filter is “samAccountName={0}”, which really is what we want. We want to query a particular samAccountName for its tokenGroups. The DOMAINUsername portion needs to be present, but the username portion is ignored. We could have used “jimmyJoeJimBob” instead of “{0}”.

The {0} represents the value collected in the “param” section of the issuance rule. I used a regular expression replacement rule to strip out the samAccountName from the UPN. Of course, I could simply have used the UPN in the query filter, which would have been cleaner. Maybe I will update this for a future project.

I had to solicit help from the Windows Higher Education list to figure this one out. There is limited documentation on the “query” portion of this rule available here. It is the only documentation I can find, and it does document the bizarre LDAP/AD query syntax discussed above:

…And some more general docs on the Claim Rule Language here:

…And an interesting use-case breakdown of the Language here:

…And a video from Microsoft’s “Identity Architects” here:

Part 6:  Where are you from?  Notes on Home Realm Discovery

When an ADFS server has multiple “Claims Provider Trusts” defined, the ADFS login page automatically will create a “WAYF”, or “Where are you from?” page to allow the user to select from multiple authentication providers.  In our case, the user would see “Active Directory” and “UVM Shibboleth”.  Since I would not want to confuse people with unnecessary choices, we can disable the display of one of these choices using PowerShell:

Set-AdfsRelyingPartyTrust -TargetName "SharePoint 2013" -ClaimsProviderName "UVM Shibboleth"

In this sample, “SharePoint 2013” is the name of the relying party defined in ADFS for which you want to set WAYF options.  “UVM Shibboleth” is the Claims Provider Trust that you want used.  This value can be provided as an array, but in this case we are going to provide only one value… the one authenticator that we want to use.  After configuring this change, ADFS logins coming from SharePoint get sent straight to Shibboleth for authentication.

Part 7: The Exciting Results

Only a sysadmin could call this exciting…

Given how heavily MS invested in implementing WS-Federation and WS-Trust into their products (MS Office support for federation services was, to the best of my knowledge, focused entirely on the WS-* protocols implemented in ADFS 1.0), I was not expecting any client beyond a web browser to work with Shibboleth.  Imagine my surprise…


IE 11 and Chrome both login using Shib with no problems.  Firefox works, but not without a glitch… upon being redirected back to SharePoint from our webauth page, we get a page full of un-interpreted html code.  Pressing “f5” to refresh clears the problem.

Office 2013 Clients:

All core Office 2013 applications appear to support opening of SharePoint documents from links in the browser.  Interestingly, it appears that Office is able to share ADFS tokens obtained by Internet Explorer, and vice versa.  The ADFS token outlives the browser session, too, so you actually have to log off of ADFS prevent token re-use.  I tested the “Export to Excel” and “Add to Outlook” options in the SharePoint ribbon, and both worked without a fuss.

Getting Office apps to open content in SharePoint directly also works, although its a bit buggy.  Sometimes our webauth login dialog does not clear cleanly after authentication.

SkyDrive Pro (the desktop version included with Office 2013) (soon to be “OneDrive for Business) also works with Shibboleth login, amazingly.  The app-store version does not work with on-premises solutions at all, so I could not test it.

Mobile Clients:

I was able to access a OneNote Notebook that I stored in SharePoint using OneNote for Android.  However, it was not easy.  OneNote for Android does not have a dialog that allow for the adding of notebooks from arbitrary network locations.  I first had to add the notebook from a copy of OneNote 2013 for Windows that was linked to my Microsoft account.  The MS account then recorded the existence of the notebook.  When I logged in to OneNote on the Android, it picked up on the SharePoint-backed notebook and I was able to connect.

The OneNote “metro” app does not appear to have the same capability as the Android app.  I cannot get it to connect to anything other than Office 365 or CIFS-backed files.

I was unable to test Office for iOS or Android because I do not own a device on which those apps are supported.

I still need to look at the “Office Document Connection” that comes with Office for the Mac, and at WebDAV clients, and perhaps some other third-party SharePoint apps to see if they work.

App-V Server Configuration, Load Balanced Configuration

In my last post I discussed issues around Kerberos configuration for an App-V 5 server cluster in a load balanced configuration.  Today I will discuss subsequent configuration requirements for making App-V publishing function in a load-balanced environment.

After configuring standalone app-V servers with Management and Publishing Server roles, I had good success with adding packages to the environment and publishing them.  However, when switching to a load balanced configuration, I experienced a failure of the publishing server to pick up on changes in the management configuration.  Helpful resources and troubleshooting notes follow:

  • A TechNet social page that I referenced in my previous post makes reference to this same problem:
    But does not point me towards any solutions.  This seems like some sort of permissions problem, so I put Sysinternals Procmon on watching w3wp.exe for “Access Denied” events, but I get nothing.  However, I do see a fair amount of database traffic at IIS startup time.
  • The following TechNet blog provided a key tipoff in App-V server diagnostics:
      The trick was to select “Show Analytic and Debug Logs” under “Actions” in the event viewer.  With this option enabled, I now see App-V management and publishing debug logs instead of just the default App-V event logs.  The debug logs contain the real error.  We see that the SID recorded for the publishing server in the management server database does not match the SID of the account making the connection!  What we needed to do was delete the publishing server entries from the management configuration, and create one new “server” under the name of the publishing server service account, not the computer account.  I just updated the SQL database entry manually, but I likely could have just used the Silverlight UI instead.  This change cleared up the mismatched SID error, but now we get an “access denied” error to the publishing metadata directory.
  • The following blog gives an excellent technical overview of App-V server infrastructure and the general troubleshooting process for resolving configuration issues:

    • Here it is suggested that I look at HKLM/Software/Microsoft/AppV/Server to review the management and publishing server configurations.  Sure enough, one problem seen here is that the publishing server is configured to connect to the management server on an http:// address.  However, I updated the management servers to use https://.  I modified those registry values and restarted IIS.  Still no luck… 
    • This blog explains how published applications are read out of a metadata xml file that is exposed to the publishing server by the management server.  Both are stored in c:programdatamicrosoftappv.  When running Procmon.exe against w3wp.exe we see “Access Denied” to these directories by our service account.  After adding “modify” rights for the service account to these directories, metadata updates again start to happen.

It is unsurprising that switching form the use of a local server account to an AD service account caused access problems for the App-V server.  The difficulty of discovering where account info and rights needed to be updated was a bit of a surprise.  But thanks to the blog-o-sphere and the mighty “procmon.exe”, we have our answers.

Now on to performance testing…

App-V 5 Server, F5 Load Balancers, and Kerberos

More fun today with Kerberos and load balancers.  Today’s challenge related to getting the Microsoft App-V publishing server to work with an F5 load balancer in a Layer 4/n-Path/DSR configuration.  Everything was working when I was accessing the individual server nodes, but when I switched to using the load balanced name and address, authentication started to fail.

After lots of log searching I eventually tried a wire trace, and found the following Kerberos error in the response from the App-V server to the App-V client:

Lots of different resources helped here:

  • This TechNet page explains various Kerberos errors and why they might occur:

    Of note is the scenario where the account handling the authentication request does not hold the SPN for which the request was made.  I set the SPN for my IIS application pool identity, but further analysis of the error packet shows that it was handled by my App-V server machine account, not the service account.  Augh!  Why?

  • This thread on TechNet Social was the biggest help:

    The user posted all of the steps they followed in configuring IIS and the service account SPN, including the tidbit:
    changed the authentication of the “Management Service” web site to useAppPoolCredentials=”true”
    I have never used this particular setting, so I dug into it…
  • The following MSDN article explains the IIS 7.0 feature of “kernel authentication”, how it affects the need for SPN entries, and its interplay with application pool identity accounts:

    Basically, with kernel-mode authentication, the SYSTEM account will handle all Kerberos authentication by default.  This explains why we were seeing Kerberos errors in the communications with the App-V client… the IIS pool identity account was not handling Kerberos delegation!

    Of special interest is this statement:
    Disable Kernel mode authentication and follow the general steps for Kerberos as in the previous IIS 6.0 version.
    [Recommended for Performance reasons]
    Let Kernel mode authentication be enabled and the Application pool’s identity be used for Kerberos ticket decryption. The only thing you need to do here is:
    1. Run the Application pool under a common custom domain account.
    2. Add this attribute “useAppPoolCredentials” in the ApplicationHost.config file.

  • This TechNet page documents how to configure Kerberos auth in IIS, and mentions the use of the IIS appcmd.exe to set the “useAppPoolCredentials” option:
    Included is the exact command line required to set the value to true:
     appcmd.exe set config -section:system.webServer/security/authentication/windowsAuthentication -useAppPoolCredentials:true
    (But the page does not really tell you what it is for, which is where the MSDN article comes in handy.)

So, Kerberos under IIS 7 and later has some nuances not present in IIS 6.  I wonder how I did not encounter this before?

SQL Reporting Services – Load Balancing Notes

As part of our evaluation of a SQL Consolidation Server, I am attempting a deployment of SQL Reporting Services in a load-balanced environment (using F5 load balancers in layer 4/fast npath config). I have followed my usual procedure for setting up a npath/direct server return config on Windows Server (with a couple of caveats… on Server 2012 the name of the MS Loopback Adapter has been changed to ‘Microsoft KM-TEST Loopback Adapter’). I then installed Reporting Services on both load balanced servers and connected them to the same back end database, and configured them to use the same SSL certificate.

Problems arise when attempting to connect to the load-balanced Reporting Services URL over SSL. Initial connection succeeds, but authentication fails. Log trawling reveals the following event:

Log Name:    System
Source:      LSA (LsaSrv)
Event ID:    6037
Level:       Warning

The program ReportingServicesService.exe, with the assigned process ID 13432, could not authenticate locally by using the target name HTTP/myserver.mydomain.edu. The target name used is not valid. A target name should refer to one of the local computer names, for example, the DNS host name.

Try a different target name.

I found a similar error discussed at the following site:

The solution there worked for me, too:

  1. Go to REGEDIT and open the key HKEY_LOCAL_MACHINESYSTEMCurrentControlSetControlLsaMSV1_0
  2. Right click MSV1_0 –> New -> Multi-String Value
  3. Type BackConnectionHostNames and click Enter.
  4. Right click on newly created value and select Modify.
  5. Enter the hostname of the site: WEBSITENAME (and on a new line enter the FQDN, WEBSITENAME.domain.com)
  6. Restart IIS

I also had considered enabling Kerberos authentication for Reporting Services, but I have decided to forgo this step (since, I think, we will want users to be able to retrieve reports from non-domain-joined computers).  However, if we did want to enable Kerberos, three things need to happen:

  1. Use SETSPN to add the HTTP/[reportServerHostName] [domain][reportServerServiceAccount]
  2. Configure the Reporting Services Service Account to be trusted for Kerberos Delegation (using AD Users and Computers, under the “Delegation” tab)
  3. Modify the rsreportserver.config file (in %ProgramFiles%Microsoft SQL ServerMSRS11.MSSQLSERVERReporting ServicesReportServer).  Find the “AuthenticationTypes” section, and specify <RSWindowsNegotiate/> as the first/only entry.
  4. Restart Reporting Services

F5 Load Balancing – Performance (Layer 4) mode and Windows Server 2008

NOTE: This article has been updated with a correction to the NetSH commands. Previously I documented the “forwarding” should be enabled on the interfaces, but “weak host receive” and “weak host send” is more accurate, as documented here:

Recently we had a problem with a web applicaiton configured for SSL-offload on our Load Balancers.  Our F5 Guru (Ben Coddington) recommended that we swich to a “Layer 4 forwarding” configuration.  In this mode, the F5 will forward TCP packets from the client directly to the web server without altering packet content, which is just what we needed.

Making this work on Server 2008 took a bit of extra leg work, though.  Here are the bones of it:

  • On the F5, create a new Virtual Server using the Type category “Performance (Layer 4)”.  Make sure that address translation and port translation are disabled.
  • Create a new F5 Pool that uses a simple port 443/ssl health monitor.  You could use any of a number of load balancing methods, but I cose “Round Robin” because it is in keeping with the “simpler is better” school of thought.
  • On the Server 2008 system, add a “loopback adapter” in the Device Manager.  (At the root of the MMC console, right-click the computer and select “Add legacy device”.  It will be of type “network adapter”, from manfacturer “Microsoft”, and have a name containing “loopback adapter”).
  • Assign the load balanced IP to the loopback adpater with netmask “”.
  • Here is the trick… you must now allow “weak host receive” on all network interfaces involved with load balancing on the Server 2008 system, and “weak host send” on the loopback interface. If this step is skipped, the Windows server will drop all packets destined for the load balancer address:
    • netsh
      interface ipv4
      set interface "Loopback Connection" weakhostreceive=enable
      set interface "Public Network" weakhostreceive=enabled
      set interface "Loopback Connection" weakhostsend=enabled
  • Make sure you have a vaild SSL certificate configured on all RDGateway systems in your farm.

That’s about it… The F5 will forward all packets sent to the load balanced IP to the next pool member in the rotation (barring persistence).  The Server 2008 host will receive the packet, and forward it to the loopback adapter (following TCP/IP routing logic).  The Server 2008 host will reply directly to the client.  Amazingly, it all seems to work.

Improving the Stability of ECTS

ECTS, the external collaboration toolkit for sharepoint, has been a bit of  a thorn in my side.  While fulfilling a vital need at our institution, it is difficult to support, and contains many bugs that apparently are not going to get fixed.

However, I recently I managed to pull off a few minor tweaks for ECTS that should improve stability in our environment.

Fix 1:  ECTSEx

First off, I installed the EXTS Extensions available in Codeplex:


My thanks to David McWee for providing these accelerators to the ECTS management interfaces.  We are now able to pull up info on ECTS users considerably more easily.  These extensions also allow password reset information to go directly to the user for whom they are intended.  Good stuff.  Unfortunately, the interface is not behaving on all nodes of our SharePoint web farm, which brings us to our next fix…

Fix 2:  Load Balancer reconfiguration

We have been experiencing intermittent problems which suggests duplication of actions by the ECTS processes.  For example, a user may receive two account activation notices, each with different passwords.  A user may change a password, only to have the new password expired immediately, followed by prompting that your password has been reset by an administrator.

My assumption is that somehow requests to either the web front end servers of the backing AD LDS instance are being replicated, resulting in replicated account management operations.  To combat this, I have reconfigured our F5 load balancers as follows:

  1. For the Big-IP Virtual Server which handles traffic to “PartnerPoint.uvm.edu” (our ECTS front-end), I set the Persistence Profile to “none”.  This prevent users from continuing to use a LB node if a more preferred node is online.
  2. For the Big-IP pool connected to PartnerPoint, we activate “Priority Group Activation” with the “Less than 1” attribute.  This will cause clients to be routed to a different Pool member priority group if there are less than one servers remaining in the highest priority group.
  3. We then set the “sharepoint3” node to be in priority group 2, and “sharepoint2” to priority group 1.  Now when users connect to the virtual server, they are routed to priority group 2 (sharepoint3), and only go to sharepoint2 if sharepoint3 is down.
  4. Repeat these three steps for the AD LDS virtual server.  This limits LDAP traffic to sharepoint3, unless sharepoint 3 is down.
  5. Now we need to limit traffic to the ECTS admin interfaces to connect preferentially against sharepoint3 as well.  We do this with iRules.  Create a new iRule which will intercept traffic to the ECTS management pages, and route it to a different pool:
    when HTTP_REQUEST {
     if {[HTTP::uri] starts_with "/ECTS/"} {
     pool WSS_partner_admin
  6. We then create a Pool with priority group activation, as outlined in steps 1-3, above, and assign this iRule to the main SharePoint virtual server.

I am able to verify though packet captures that all requests to the ECTS admin pages, and all traffic to the ECTS front ends is being routed preferentially to “sharepoint3”.  So far, this seems to have helped.  If nothing else, the admin pages are now loading consistently, which was not true for us in the past.