Im setting up three servers with DFS, it is Image servers attached to webservers, where the images is displayed. It's about 350gb of pictures on each server, but they are slightly out of sync.

Now when I want to setup DFS replication between the three of those, I've chosen a primary member. The question is then:

Should I set up the primary server (Most up-to-date) as the only server on which the pictures is written, so the replication process is then replicating the pictures to the two other servers? I guess if I still was writing all pictures to all three of the servers, there will be a lot of Conflict And Deleted.

The whole question is, what is happening if something is uploaded to the primary server while the initial replication process is running?


I have a similar setup, but with more servers in my DFS replicated list. They are image servers as well. I have them set up to all replicate, instead of the hub and spoke I did the spiderweb approach - HOWEVER - the code is written in such a way that it tries to drop the image on server 1, if it succeeds it stops there. If for some reason the save to image 1 fails it saves to image 2, then 3 etc etc. This means if I want to take my servers down to do updates or other maintenance I don't have to worry about things not working in the mean time.

Hope that helps.

  • This was actually more about the initial replication process, what happens if files are written to the primary server while still in the initial replication process. I did a little test with out 1gb of files, one server had 600mb of files, the other had 800. I made the 800mb server the primary and started replication. Once the second server picked up the replication (Had replicated about 100mb of data) I c/p a folder consisting of data equal to 200mb to the primary server, and it was replicated just fine.. I guess it somehow answers my question. – psalomonsen Aug 6 '10 at 6:25
  • Im thinking. Your setup and all, looks similiar. But, what server are you (after your initial replication) writing your images to? Do you write to the namespace? (I know the namespace is the virtual path to all X numbers of folders, in my case 3) So, if you write to the namespace, it only saves on one server, and then replicated to the other? – psalomonsen Aug 6 '10 at 6:27
  • Copying in a file during "initial replication" is handled the same way as copying a file in after it's done. The way you should look at it is you're turning on replication, it sees everything in that folder as "new stuff to replicate". Copying in more items is just "more new stuff to replicate". Like I said our code is attempting to write to the replicated folder on server 1, if it fails it moves to server 2 and down the road until it finds a place it can write to. I haven't ever played with writing to the virtual path namespace. – Dave Holland Aug 6 '10 at 14:40
  • What happens if you have a big gap in data, and you put one, lets say server2 as master server in the initial replication, and someone uploads a new file, to either server1 or server3, will it go to PreExisting, or will it actually handle it as a new file? Or can you only write to the master server in the initial replication process? – psalomonsen Nov 3 '10 at 14:45
  • DFS should be smart enough to see new files as new files - even during the initial replication. I haven't actually tested that - but I can tell you that when the file is placed in the directory isn't the only time DFS checks to see if everything is current. If you're in doubt you can always restart the DFS service and it will re-check the distributed directories. – Dave Holland Nov 4 '10 at 5:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.