Yes, I left out the AutoDiscoverInternalUri piece and it's defiantly important, and we set that up as well, of course. On Feb 25, 2015 11:36 PM, "Adam Farage" <[email protected]> wrote:
> The description below is a right but missing a key point. > > In the event the client is domain joined (client being a computer running > Outlook that is compatible with the version of Exchange 2010 in the > environment) it actually looks at the SCP (Service Connection Point) within > the local Active Directory sites CAS first. To find this within Exchange > Management Shell its a simple command: > > *Get-ClientAccessServer | Select Site, AutoDiscoverInternalUri* > > Depending on how you are setup (either active / active or active / > passive, as I saw you said it is spread across two sites) you would then do > the DNS records as shown below. I would recommend just setting up a DNS A > record to the Virtual IP of the load balancer for the CAS Array (if you > have one) or have it go to an individual CAS if you are not doing load > balancing. The InternalURL and ExternalURL of the AutoDiscover Virtual > Directory are not used and exist due to the formatting of the virtual > directory object within Exchange / AD Schema :) Nothing is actually > referenced to it or anything.. yay Microsoft. > > As for the comment below about creating a new AD site for the CAS Array, > don't. Technically the only function of the CAS Array address is to point > to the MSExchangeRPCClientAccess attribute on the mailbox database (if the > mailbox databases are created before the CAS Array is implemented you need > to change that, if its afterwards it will auto-populate if I recall). When > you create a CAS Array it is setup for the local AD site that you declared > within the command *New-ClientAccessArray*. From there *all* Client > Access Servers that are then implemented into that (AD) site are apart of > this CAS Array. You also cannot create a CAS Array without specifying the > site. > > There is no problem creating two CAS Arrays as its just a logical AD > object. I would keep the AD Sites geographical to there proper location / > subnet, and then do load balancing (L7 if its Exchange 2010, and I will > still recommend L7 for Exchange 2013 for most situations although that is a > separate conversation) if you have more than two CAS. As others pointed out > you could.. in theory use Windows NLB but I would be careful. Some of the > limitations is that you cannot have a DAG member + CAS running Windows NLB > on the same server (due to the cluster DLLs and Windows NLB DLLs > conflicting), and it is not recommended to scale past 7 CAS. You also have > to enable multicast I think (its been a very long time since I have touched > Windows NLB - I typically recommend KEMP virtual appliances or hardware. > Cheap and does the job). > > And to make things more fun, Exchange 2013 has no CAS Array object (and > most were confused on what the logical object did exactly which is why > Brian Day wrote that article). > ------------------------------ > From: [email protected] > Date: Wed, 25 Feb 2015 21:58:19 -0500 > Subject: Re: [Exchange] CAS Array question > To: [email protected] > > It is very incorrect that Outlook doesn't update when there is a CAS > server migration, as long as Autodiscover is working properly in the > domain. > > The best way we setup Autodiscover was as an alias that resolved to the IP > address of an Exchange aware Load Balancer. With EX2010, you can use the > free Windows Network Load Balancer IF you have your mailbox and CAS roles > on separate servers. If you have everything in one box, then you should buy > a 3rd party load balancer, This is all the recommendation of Microsoft, > not me. They certify many Load Balancers to use, we use Kemp and are happy > with them. > > So, for example, in our environment: > > In DNS, created an A record with IP address of load balancer called > mail.contoso.com > In DNS, created an Alias record for Autodiscover.contoso.com that points > to mail.contoso.com > > On Load Balancer, created a Virtual Service with the IP corresponding to > mail.contoso.com. In the settings for the virtual service, listed the > real server IPs of the CAS servers. > > We just finished an upgrade from EX2010 to EX2013 (so using new servers) > using this method and never had to update anyone's Outlook. We never had a > CAS array either before. > > With the project you are facing, and if I were a consultant, I would > suggest you upgrade to EX2013 because it is much more friendly with DAG and > CAS servers in different AD sites. Also, there is no longer a need to Load > Balance at layer 7 as with EX2010, which makes CAS load balancing much less > of a pain the butt. In fact, you can even use DNS round robin for load > balancing in EX2013, but you will not see anyone recommending that approach > for medium to large organizations. > > > > > On Wed, Feb 25, 2015 at 3:51 PM, Kennedy, Jim < > [email protected]> wrote: > > > http://blogs.technet.com/b/exchange/archive/2012/03/23/demystifying-the-cas-array-object-part-1.aspx > > > > > > *From:* [email protected] [mailto: > [email protected]] *On Behalf Of *Richard Stovall > *Sent:* Wednesday, February 25, 2015 3:47 PM > *To:* [email protected] > *Subject:* Re: [Exchange] CAS Array question > > > > The CAS array is apparently just an AD object that tells internal Outlook > clients where to find CAS servers. If you don't create one at Exchange > 2010 installation, one is created for you with the fqdn of your mailbox > server. The issue is down the line (like now, for me) when you want to > move mailboxes to a new server in order to decommission the old one. > Outlook doesn't update and each user profile has to be reconfigured either > manually or via pushing a prf file. (I may have some of the description > not quite right, but that's the gist of it.) I figure it's better to fix > the problem now rather than leave it to the next person to figure out. > > > > On Wed, Feb 25, 2015 at 3:37 PM, Steve Ens <[email protected]> wrote: > > I dunno Richie, do you really need an array for 100 users? I have a > single server virtualized. Never any problems with client access. > > > > On Wed, Feb 25, 2015 at 2:29 PM, Richard Stovall <[email protected]> > wrote: > > I will soon be moving Exchange 2010 to new hardware (or maybe virtual > hardware, but that's another question). > > > > I have a single server with fewer than 100 users spread across two AD > sites. Of course I did not create a CAS array all those years ago, so I am > preparing to deal with the ramifications of that now. > > > > Currently all mailboxes are (obviously) hosted on one server in one of the > sites, but that may change in the future if we grow, add data center > capacity, whatever. > > > > Given the need to create a CAS array now, should I create a single > site-specific array for each site? Can an array for site 2 specify an ip > in site 1? If not, should I create a single array and not specify the site? > > > > Thanks for the help, > > > > Richard > > > > > > >
