DFS replica set failover

[Chris Turner]

Hi,

Situation:
I have a Win2K file server that hosts data in multiple
user shares on it's d: partition which is running out of
space. We bought another disk to add as another partition,
f:.
The goal here is to migrate the user's shared data on d:
to f:, freeing up space on d:, with minimal/no disruption
to the users.
DFS seems to offer a solution to this problem. The idea is
to replicate the data from d: across to another share on
f: & then 'switch off' the d: share - hey presto!

Problem:
I realise I would have to implement the use of the DFS
tree rather than direct shares to be able to take
advantage of its failover functionality. This would not be
a problem, but in testing the idea, I cannot see a way
to 'switch off' the old linked share on d: in the DFS so
that users stop saving data to it & automatically move
over to the new linked share on f:.

I've tried using the 'take replica offline/online' flag in
the DFS manager, but this does not seem to force users
over to the replica share as I was expecting & data is
still saved to the old share.

Any ideas on how to do this/what I'm doing wrong?

 

[Daniel Billingsley]

Taking it offline will work, but the effect is not immediate. Once a client
gets a reference to one target or the other it will just cache that
reference for a period of time.

If you can, take the d: target offline and reboot the server after hours and
you'll be golden. Or everyone should eventually switch over when their d:
references timeout - I don't remember the default TTL for a dfs reference
off the top of my head.

 

[Chris Turner]

Thanks for the response, Daniel.

I think I've realised I'm being a dunce anyway because I
was testing the failover between two machines instead of
between two partitions on the same machine, which doesn't
seem to be allowed. Does anyone know if there is a way to
do this?

I may try replicating it over to a different 'surrogate'
server & then back to the second partition to see if that
works.

Cheers.

 

[Daniel Billingsley]

I've done that last idea myself so it should work no problems.

Is the amount of data very large? You may need to check into staging area
and journal wrap issues in the KB to make sure you have things set up
properly before you start on the real data.

 

[Daniel Billingsley]

You can also find good coverage of those topics in the win2k3 documentation
in case you're using that.

 

[Mukul Gupta [MSFT]]

Clients could use 'dfsutil.exe /pktflush' on client machines to force them
to get new DFS referrals instead of waiting for TTL to expire on the cached
entries. (dfsutil.exe is reskit/support tool)

--
Thanks
Mukul [MSFT]

PS: Please post DFS related queries in newsgroup
microsoft.public.server.dfs_frs. Please use "DFS" in subject to make it
immediately noticeable.

PS: Please find latest dfsutil.exe for Win2003 (updated) at
http://www.microsoft.com/downloads/...C3-C7A3-4B5B-9FF1-FD01B0203DC1&displaylang=en

Disclaimer: This posting is provided "AS IS" with no warranties, and confers
no rights