Windows Server

Fab · Guest

Hi there,

I have a Dfs setup that consists of 12 replica sets each of which with
50 GB or more data. I've had this system running for almost a year now,
everything was running smoothly until recently that the computer
hosting the source files ran out of space on the drive configured as
FRS staging area. As a result of that, FRS has gone into error state
complaining about a JournalIDMismatch error which probably means a
journal wrap.

Since each of my replica sets has two partners, the source and the
target, I am left with two options:

Option 1. Perform a D2 with the following details:
On the source computer:
a) Add a bigger drive to make the staging area bigger
b) Relocate ntfrs.jdb to force it to re-initialize with all the files
and folders that sit on the FRS shares

And, then perform a D2 on the replication partner.

Option 2. Backup and Restore:
a) Add a bigger drive to the source computer to make the staging area
bigger
b) backup the data on each FRS share of the source computer
c) Restore the backup data on the replication partner
d) turn on the replication for the replica set

Repeat step a-d for each of the replica sets. The only glitch here is
that since each FRS shares contains about 50 GB, I'll have to do the
backup over the network. Now, the question is if there is any benefit
in choosing option 2 over option 1? Will it save me time or network
bandwidth?

I'll appreicate your comment/feedback.

thanks,

Fab

Paul Williams [MVP] · Guest

There's not going to be any difference unless you backup NTFRS.JDB. Without
that file, FRS will replicate all files again as there are no replication
checkpoints.

Normally I would recommend a D2. However with that much data your WAN, USN
Journal and Staging areas are going to get hammered (and possibly wrap
again). I would try the backup restore option on the server that is in an
erroneous state -just make sure you are backing up the JDB as well.

As a for info. (although it's no help now), the new DFSR feature that is
coming in R2 will solve this problem through it's "likeness" technology
(can't remember the correct terminology, but believe that's what the DFSR PM
called it - might have been "similar" ;-)...

--
Paul Williams
Microsoft MVP - Windows Server - Directory Services
http://www.msresource.net | http://forums.msresource.net

Fab · Guest

Thanks Paul. My issue is that it is actually the FRS on the (source)
server that has gone into error state which means ntfrs.jdb on this
server needs to be re-initialized. So, backing up the old ntfrs.jdb is
not going to help me as it's erroneous state. Perhaps, my best bet
would be to
1. stop ntfrs on both replication partners
2. Set the USN Change Journal size registry entry to a large value
(e.g. 4 GB)
3. delete ntfrs.jdb on both replication partners
4. Enable replication policy but do it with one replica set at a time.
This way, I won't flood the network and there probably is less chance
of another joural wrap.

Another thought is to treat this as a new Dfs setup where the source
folders already contain data. Then, I can proceed with the pre-staging
approach and setup the Dfs that. Do you think if it will make a
difference? What are the trade-offs?

Another question:
Let's assume I am setting up a Dfs link where the source folder already
contains some data. I realize replication doesn't start until I set the
configuration policy, but the question is "when does it start
replicating the first file"? Is it the case that ntfrs performs a
complete scan of the share (to get a complete view of "all existing"
files) and builds the ntfrs.jdb database before it starts sending
notification across to the replication partner? Or, it is rather the
case that the FRS on the source computer scans some of the files,
builds ntfrs.jdb, sends notification to the partner and while FRS of
the partner computer is fetching the first set of files, FRS on the
source computer will continue building ntfrs.jdb for the remaining
files?

Your thoughts/comments will be greatly appreciated.

thanks,

Fab

Paul Williams [MVP] · Guest

Hi Fab, sorry for the delay!

If this is just one, then you could just D2.

Although, if you can use a DR/ remote initial setup:

	Windows Server Forum Index -> DFS and FRS	All times are GMT
Page 1 of 1