MOSS Small Farm Installation and Configuration War Story

This week, I’ve struggled a bit with my team to get MOSS installed in a simple two-server farm.  Having gone through it, I have a greater appreciation for the kinds of problems people report on the MSDN forums and elsewhere.

The final farm configuration:

  • SQL/Index/Intranet WFE inside the firewall.
  • WFE in the DMZ.
  • Some kind of firewall between the DMZ and the internal server.

Before we started the project, we let the client know which ports needed to be open.  During the give and take, back and forth over that, we never explicitly said two important things:

  1. SSL means you need a certificate.
  2. The DMZ server must be part of a domain. 

Day one, we showed up to install MOSS and learned that the domain accounts for database and MOSS hadn’t been created.  To move things along, we went ahead and installed everything with a local account on the intranet server. 

At this point, we discovered the confusion over the SSL certificate and, sadly, decided to have our infrastructure guy come back later that week to continue installing the DMZ server.   In the mean time, we solution architects moved ahead with the business stuff.

A weekend goes by and the client obtains the certificate.

Our infrastructure guy shows up and discovers that the DMZ server is not joined to any domain (either a perimeter domain with limited trust or the intranet domain).  We wasted nearly a 1/2 day on that.  If we hadn’t let the missing SSL certificate bog us down, we would have discovered this earlier.  Oh well….

Another day passes and the various security committees, interested parties and (not so) innocent bystanders all agree that it’s OK to join the DMZ server with the intranet domain (this is a POC, after all, not a production solution).

Infrastructure guy comes in to wrap things up.  This time we successfully pass through the the modern-day gauntlet affectionately known as the "SharePoint Configuration Wizard."  We have a peek in central administration and … yee haw! … DMZ server is listed in the farm.  We look a little closer and realize we broke open the Champaign a mite bit early.  WSS services is stuck in a "starting" status. 

Long story short, it turns out that we forgot to change the identity of the service account via central administration from the original local account to the new domain account.  We did that, re-ran the configuration wizard and voila!  We were in business.

</end>

Subscribe to my blog.

Technorati Tags:

5 thoughts on “MOSS Small Farm Installation and Configuration War Story

  1. Cimares
    It’s perfectly ok to have your SQL in a different Vlan/subnet than your WFEs. In fact it’s recommended, after all as mentioned before, what Security expert is going to let you stick SQL in the dmz? The recommendation is that your SQL traffic does NOT use the same interface cards as the user traffic, however even this connection may pas through a firewall for additional protection.
     
    The restriction related to multiple WFEs in a farm environment relates to if you’re using Microsoft load balancing, then these must all be in the same VLan.
    Reply
  2. Paul

    I can almost beat your SSL certificate issue.  We had everything created and were ready to extend the web app with SSL (then redirect port 80 in IIS).  The administrator had a .cer file ready to go.  But NONE of the options or crazy contortions to apply it in IIS will work–the site always displays a blank page like the site collection doesn’t exist.

    After much banging of heads, we learned this was caused by the cert request not coming from that server.  The administrator simply asked for a cert and was emailed the resulting key.  With no private key, the SSL tunnel could not get built between the WFE and the browser.  We wasted 1/2 day on that.

    Reply
  3. Christian wrote:
    Very interesting! I highly doubt that it shouldn’t be supported to host the WFE’s in one VLAN/DMZ and APP/SQL in another VLAN/DMZ.
     
    The TechNet articles about supported Extranet scenarios doesn’t have any reservations, either – but TechNet could be incorrect 🙂 None of our clients would allow their SQL Servers to sit on the same VLAN/DMZ as the WFE, so I sincerely hope the MS got it wrong.
     
    Can you elaborate on what should be the problem with spitting the configuration? Performance reasons only? Or do they in fact mean that the WFE’s should be on the same VLAN/DMZ? That would make more sense to me.
     
    Sincerely,
    Christian
    Reply
  4. Paul Galvin
    That’s a very good question.
     
    We are tracking very closely to the MS documentation, so I can’t imagine how they would refuse to support it.  That said, I am not an infrastructure person, so it’s possible that I’m abusing terms in my post.
     
    As I understand it, the correct approach is to have (at least) two AD domains.  One internal domain and one in the perimeter network.  The perimeter network’s AD would have a "limited trust" relationship with the internal AD.
     
    But you probably already know all that 🙂
     
    Bottom line, I don’t know.  We did not receive or look directly to Microsoft for guidance on this one.
     
    –Paul G
     
    Reply
  5. Tom Dietz
    Is this configuration supported?  At the SharePoint Conference in Seattle in March, I was chatting with some Microsoft Engineers and they said that supported configurations do not allow WFEs to cross VLANs or routers.  I assume that since the WFE is in a DMZ, it is crossing some sort of firewall/router or is in its own VLAN. 
     
    So basically the DB and WFE/App Servers all have to be on the same VLAN.
     
    They were really adamant about this–it’s actually a slide in the ‘Geographical’ deployment session if you have access to the deck.
     
    I’ve read TechNet articles that illustrate sample configurations that contradict their statements, but the MS guys basically said that TechNet is wrong.
     
     
    Reply

Leave a Reply to Paul Galvin Cancel reply

Your email address will not be published. Required fields are marked *