Disaster Recovery with DFS

[RJ]

Hi

We have, say, 3 servers - A,B and C. B and C replicate their data to
server A via DFS, so all data is on server A - and as such, server A
is the only server that needs to have data backed up.

(B and C back up C: and system state only)

So - assume server B dies and needs a rebuild.....

I KNOW we can build new server, DR server B C: and system state, and
create file tree for DFS on data volume, wait for X hours, and all
data is replicated by DFS and everything is perfect! Woopee!

However, now assume server B is over a 2Mb link and has 30Gb of data,
and clearly we don't want to replicate this over the WAN via DFS.

I understand we can restore the data area of server A (that was
replicated) to server B to server B (using Windows 2003 compatible
backup program) - and then move folders to appropriate loction and
start DFS - and again - DFS sorts it out. (e.g. drive tape to remote
site and restore locally).

What we can't prove in the test is whether data is replicated or
moved/copied from the "preexisting" folder in each DFS share! In the
test lab with full DR replicas of the environment, server A has to be
brought up as "authoritave for all replicated data" (e.g. AD) in order
for AD to work - as there are no other "real" DCs in the test lab -so
we believe this makes it teh master for DFS too - and hence test is
not real.

Can't find any white papers or help on this topic - except "let it
sort itself out" - which is fine - but time consuming with lots of
data over WAN.

Ideas/thoughts most welcome! Any questions, reply here and I'll get
back to you.

THANKS!

 

[Jill Zoeller [MSFT]]

For FRS, if prestaging is done correctly, you will see the preexisting
folder shrink in size as the files are moved locally. You can also use a
netmon trace to confirm that data is not being replicated across the
network. If your WAN is slow, the prestaging would be much faster, too.

 

[RJ]

Jill - thanks.

Any papers on "how" to prestage correctly? I am assuming when AD is
up, make "dead replica" unavaialble. Restore data to dead server
(when rebuilt!), stop FRS, move folder tree to correct folders, start
FRS and then make replica avaialble again.

Is this right?

If so - great - but one final question:-

In a DR lab, if server A is DR'd as an authrative master for all
replicated data, does this change things for FRS/DFS too? Happy for
it to be a master as long as it doesn't want to push all data back to
"slave" replica.

Thanks