Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I'm considering setting up a way to automate data redundancy and synchronisation of multiple workstations at the office. The way our office is setup, nobody has a fixed desk and anyone can login and work on any workstation on the network.

One way of doing this would be to have a NFS mounted /home filesystem. However, that would slow down file access as everything would need to be pulled off the network plus it creates a single point of failure.

This is the general idea.

[WS-A]--- (DRBD) ---[BACKUP]--- (DRBD) ---[WS-B/C/D/etc]

The idea is for someone to work on WS-A and whatever file saved is automatically synchronised with another machine (backup). So, the data is now available on more than one machine.

Ideally, the data should synchronise and propagate itself across all the workstations WS-B, WS-C, etc on the network. This propagation and synchronisation does not need to be instantaneous and can be asynchronous as long as the same user can log into any workstation at another time (but only one machine at a time) and find their latest files present.

Also, if any workstation crashes, the setup will automatically recover and resync after the machine comes back onto the network, preventing split-brain problems.

  1. Would this even be possible with DRBD; and
  2. Are there any nightmare synchronisation issues that we might need to worry about?

PS: We currently have a similar setup that uses glusterfs and it seems to work thus far, except that the performance drops when accessing a large git repository over it.

share|improve this question
2  
O HAI! NFS calling! –  Tom O'Connor Feb 28 '12 at 17:04

2 Answers 2

up vote 2 down vote accepted

Putting workstations into a DRBD cluster sounds like a terrible idea to me. I doubt that this will ever work reliable because now users can't restart their workstation anymore.

Instead, make sure that your NFS server isn't a single point of failure anymore and setup a HA system at this point. Since this is not really easy, I just link to a tutorial for this.

share|improve this answer
    
why wouldn't they be allowed to restart their machines? wouldn't that just require a resync as if the machine went offline and then came back online later? –  sybreon Feb 29 '12 at 0:29

What you are trying to setup reminds me of the way Windows roving profiles work.

  1. Get profile from server to local disk
  2. Upon logout sync profile from local disk back to server

Basically you could achieve the same using rsync during login and logout from/to a central server (which should be HA, of course).

IMHO DRBD-configuration is a little bit too static for your purpose.

With DRBD you will run into trouble if you have a single user with multiple logins. If you use rsync -e ssh -auHS --delete the last update will win and the latest delete before logout will propagate.

BTW: Doing the sync via ssh will make all this secure, too.

share|improve this answer
    
yea, this was considered but the rsync method would require that the changes are 'master' on the central server. also, the logout/login rsync might take a long time depending on the amount of changes. it might not completely sync either during shutdown. –  sybreon Feb 29 '12 at 0:27

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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