DFS file question

Associate
Joined
8 Mar 2007
Posts
2,176
Location
between here and there
hey guys,

I currently have my network shares like this;

iscsi target on nas
- connected to DC2

DFS namespace on dc2 configured like this;

namespace (\\domain.com\south) -> add folder (iscsi target F:\ - users)

this allows me to access the files on the NAS using the namepsace ie, \\domain.com\south\users.

What I really want to do is have it so that if dc2 fails, dc1 picks up the slack.

How can I go about this without using DFS replication, as my understanding is I need twice the storage space, one for live and one for the replicated data. (which i don't have)

Can I simply connect my dc1 to the iscsi target, then somehow add that to the DFS namespace, so if one isn't available my users get the other?

This of course will only work if if the nas hasn't failed.

any help, pointers, links greatly appreciated;

thanks in advance :)
 
Associate
Joined
18 Sep 2012
Posts
2,212
You will need to setup Microsoft Cluster Services and share the DFS folder(s) from the cluster (which can fail over between DC1 and DC2, but is only ever on one of those at any given time).
 
Associate
OP
Joined
8 Mar 2007
Posts
2,176
Location
between here and there
arh, not going to be able to do that then.

Server 2008 std on both dc's.

I'll install the dfs role service and get the namespace on, but it'll have to be a manual intervention to connect the iscsi target if the dc 2 fails.

thanks for getting back to me.
 
Associate
Joined
1 Dec 2005
Posts
803
What happens if/when your iSCSI target goes down? Seems a bit pointless bothering with DFS when your storage has no redundancy... ;)
 
Associate
OP
Joined
8 Mar 2007
Posts
2,176
Location
between here and there
What happens if/when your iSCSI target goes down? Seems a bit pointless bothering with DFS when your storage has no redundancy... ;)

because i don't have a spare 2tb nas kicking about.

you can spend your entire life building in redundancy. the difficult bit is getting the best that you can with the tools you have provided.
 
Associate
Joined
18 Sep 2012
Posts
2,212
You would be much better off getting rid of the iSCSI, installing local storage on both DC1 and DC2, and using DFS-R to replicate between them. That would give you true redundancy across everything (servers, storage, network), *and* you don't need Enterprise Edition.
 
Associate
OP
Joined
8 Mar 2007
Posts
2,176
Location
between here and there
You would be much better off getting rid of the iSCSI, installing local storage on both DC1 and DC2, and using DFS-R to replicate between them. That would give you true redundancy across everything (servers, storage, network), *and* you don't need Enterprise Edition.

true, but I'll have to go buy 4 2tb drives to stick 2 in each, then configure raid mirrored, then dfs replication... blah blah blah...

not a bad way to go. I'll fire up Oracle later and see what in stock...

cheers
 
Soldato
Joined
7 Jun 2003
Posts
16,032
Location
Engerland
You can have multiple name space servers allocated to a single folder target, is that what you want?

EDIT: Wonderful thing about iSCSI, you can map storage to two locations at once for the same storage unit, just connect the same storage to both servers and don't enable replication but add both servers folder targets. You get the benefit of connectivity redundancy but without the storage redundancy
 
Last edited:
Associate
OP
Joined
8 Mar 2007
Posts
2,176
Location
between here and there
EDIT: Wonderful thing about iSCSI, you can map storage to two locations at once for the same storage unit, just connect the same storage to both servers and don't enable replication but add both servers folder targets. You get the benefit of connectivity redundancy but without the storage redundancy

not true (in my experience)

I tried this once before and killed all the data on the disks..... didn't try it again.. got a 'right royal telling off'
 
Associate
Joined
18 Sep 2012
Posts
2,212
mrbios: the only way two hosts can talk to the same iSCSI LUN without corruption is if there is some sort of cluster-aware file system in place. In the case of NTFS you need Microsoft Cluster Services, which isn't available in the Standard edition of Windows 2008.
 
Associate
Joined
1 Dec 2005
Posts
803
You would be much better off getting rid of the iSCSI, installing local storage on both DC1 and DC2, and using DFS-R to replicate between them. That would give you true redundancy across everything (servers, storage, network), *and* you don't need Enterprise Edition.
This. It's exactly what I've setup at home (!) and works a treat. My only word of caution would be to run DC2 as read-only, basically one-way replication. I've had problems with things like Premiere Pro getting confused and trying to write to files on both targets, DFSR getting very confused and deleting files when it tries to merge the changes. Bad times.

There's probably a better way around it but for me that solution works.
 
Associate
OP
Joined
8 Mar 2007
Posts
2,176
Location
between here and there
I've been chatting with some guys from other companies I know and they're telling me that if your talking about large amount of data that is in constant use, this isn't the way forward.

they've experienced issues with old files being displayed and what not. the idea of making in one way is quite good. That could be revoked if the other DC dies..
 
Associate
Joined
1 Dec 2005
Posts
803
Yes large files can be tricky, and/or a large amount of data that needs replicating. The key is ensuring the staging folders are set to an appropriate size but this can end up being quite a sizeable amount of space. Transfer speed is normally good though once everything is configured properly.
 
Soldato
Joined
5 Jun 2008
Posts
6,238
Location
Portsmouth/Fareham
You would be much better off getting rid of the iSCSI, installing local storage on both DC1 and DC2, and using DFS-R to replicate between them. That would give you true redundancy across everything (servers, storage, network), *and* you don't need Enterprise Edition.

This, although I wouldn't attach the storage to the DC's unless they are your only option.

Ensure the staging quota is large enough (it needs to be big enough to cover your largest file in the store really, article here http://technet.microsoft.com/en-us/library/cc754229.aspx)

Replication shouldn't be too much of an issue, just set it to bi-directional so that what happens to one, happens to the other. Some users will be targeting the different folder target servers and as a result they both need to be syncing.
 
Top Bottom