Server 2003 R2 - DFS & failover

[oldtechie]

Greetings to all,

Distributed File System Technology Center

http://www.microsoft.com/windowsserver2003...fs/default.mspx

Scroll down to Introductory Overviews. Just below that is a link 'Branch Office Demo'.

There is a video that runs about 14 minutes. Please play the video, and be aware towards the end where he says the London server fails, but the user(s) in London will still be able to access their files via the Redmond server.

How are the London user(s) going to physically get past a failed server unless they are using an internet connection in the first place?

I think he just assumes we know he is talking about WEB servers. But he does NOT say that.

Could someone please clarify this for me?

Edited April 7, 2009 by oldtechie

[cluberti]

I am taking a course on Operating Systems. In the textbook they mention DFS failover/failback. I am confused because they do not mention how the clients are accessing the server. If clients were going through a router and it were connected to the server via ethernet it seems to me that a total failure of the server would render all clients without service.

DFS is a mirrored file system, fronted by a "virtual" server name that points you to (hopefully, if configured properly) the closest replica of the DFS to your location. However, because all servers in the grouping are aware of whether or not other replicas are down (similar to heartbeats in clusters or web farms), and you're never accessing a DFS replica server directly (you access it via the DNS DFS namespace root name in AD or in the root of the non-AD-integrated DFS) when you go to hit a server you are directed to another replica if the closest to yours is down. Obviously there can be issues with data availability on a remote replica depending on whether or not what you have on the DFS share has been replicated, but the article is correct, if one node is down you will be re-routed to another without you really knowing it without looking at network traffic.

[oldtechie]

Thanks Cluberti,

You commented on the first 5 lines. I edited that part out..

Sorry it only adds confusion. My real question is on down at the bottom.

This has to be a simple question.... Please, whoever reads this, I can and have read quite a bit on the Microsoft Web Sites. I am just confused on one little issue.

I don't understand how the CLIENTS are (physically) accessing the server(s). Once I understand that, all the rest will make sense to me.

Later today......

OK.. I just phoned a guy in my local computer club. The answer seems to be that the clients connect through the Internet to the server(s). That is what I needed to know.

Edited April 7, 2009 by oldtechie

[fizban2]

old techie,

you are right, a client who has no LOCAL dfs nodes will traverse your companies intranet to find the next closest node. it doesn't out onto the internet persay as much as it will use a dedicated link between locations to find the next closest replication point. just wanted to point out that using DFS across the internet is not something you do or would want to do. dedicated links between sites are the way to go