Discussion:
DFS on Windows Cluster
(too old to reply)
MarkT
2007-03-28 00:04:15 UTC
Permalink
We are wanting to implement Windows Clustering in two separate data centers
for the sole reason of hosting Windows 2003 R2 DFS. As I understand,
domain-based namespace is not supported on a cluster but a stand-alone
namespace is supported.

I am not clear on the following statement from The Step-by-Step Guide for
DFS
:(http://technet2.microsoft.com/WindowsServer/en/library/14211e1f-2dbd-418d-b721-2005c725ce351033.mspx?mfr=true )

"On server clusters, replicated folders should be located in the local
storage of a node, because the DFS Replication service is not cluster aware
and the service will not fail over to another node."

1 - Can you clarify the above statement?
2 - Can you provide any resources for configuring DFS on a Windows Cluster?
3 - Other than the standard limits of DFS might you have any other
recommendations?
Jill Zoeller [MSFT]
2007-03-28 16:58:44 UTC
Permalink
DFS Replication is the replication component, not the namespace component
that you refer to for stand-alone namespaces. What this means is that you
can't use DFSR to replicate content stored on shared cluster storage.

Can you clarify if you are also interested in replication using DFSR? I
can't quite tell if you are asking for DFSN (namespace) guidance or DFSR
(replication) guidance.
--
This posting is provided "AS IS" with no warranties, and confers no rights.

Want to learn more about Windows file and storage technologies? Visit our
team blog at http://blogs.technet.com/filecab/default.aspx.
Post by MarkT
We are wanting to implement Windows Clustering in two separate data centers
for the sole reason of hosting Windows 2003 R2 DFS. As I understand,
domain-based namespace is not supported on a cluster but a stand-alone
namespace is supported.
I am not clear on the following statement from The Step-by-Step Guide for
DFS
:(http://technet2.microsoft.com/WindowsServer/en/library/14211e1f-2dbd-418d-b721-2005c725ce351033.mspx?mfr=true
)
"On server clusters, replicated folders should be located in the local
storage of a node, because the DFS Replication service is not cluster aware
and the service will not fail over to another node."
1 - Can you clarify the above statement?
2 - Can you provide any resources for configuring DFS on a Windows Cluster?
3 - Other than the standard limits of DFS might you have any other
recommendations?
MarkT
2007-03-28 17:50:04 UTC
Permalink
The request is for us to create a new DFS Namespace on a (2) node cluster.
The namespace would need to be a stand-alone DFS namespace because the
domain-based DFS namespace is not supporte on a cluster.

All the data would then be copied to the new DFS Namespace. Next, we would
build a second (2) node cluster and configure it as an additional server for
the namespace then use DFS Replication between the two nodes. But from what
your saying, the DFS Replication service is not cluster aware so this is not
a supported configuration.

So a standalone DFS Namespace is supported on a cluster but DFS Replication
is not. Is there any reason at all to put a DFS Namespace on a Windows
Cluster, as suggested in the DFS:FAQ
http://www.microsoft.com/windowsserver2003/techinfo/overview/dfsfaq.mspx
Post by Jill Zoeller [MSFT]
DFS Replication is the replication component, not the namespace component
that you refer to for stand-alone namespaces. What this means is that you
can't use DFSR to replicate content stored on shared cluster storage.
Can you clarify if you are also interested in replication using DFSR? I
can't quite tell if you are asking for DFSN (namespace) guidance or DFSR
(replication) guidance.
--
This posting is provided "AS IS" with no warranties, and confers no rights.
Want to learn more about Windows file and storage technologies? Visit our
team blog at http://blogs.technet.com/filecab/default.aspx.
Post by MarkT
We are wanting to implement Windows Clustering in two separate data centers
for the sole reason of hosting Windows 2003 R2 DFS. As I understand,
domain-based namespace is not supported on a cluster but a stand-alone
namespace is supported.
I am not clear on the following statement from The Step-by-Step Guide for
DFS
:(http://technet2.microsoft.com/WindowsServer/en/library/14211e1f-2dbd-418d-b721-2005c725ce351033.mspx?mfr=true
)
"On server clusters, replicated folders should be located in the local
storage of a node, because the DFS Replication service is not cluster aware
and the service will not fail over to another node."
1 - Can you clarify the above statement?
2 - Can you provide any resources for configuring DFS on a Windows Cluster?
3 - Other than the standard limits of DFS might you have any other
recommendations?
Jill Zoeller [MSFT]
2007-03-28 20:27:52 UTC
Permalink
Remember there are two things you want to have redundant here: (1) the
namespace and (2) the user data. These are both separate technologically.
They don't even have to exist on the same server.

The namespace has a service component that hands out referrals to clients.
If you want users to be able to access the namespace (i.e., receive
referrals), you need to make sure that the service can continue to run if
one of the namespace servers (aka root servers) goes down. One way to do
this is to host a stand-alone DFS namespace on a server cluster. Another way
is to create a domain-based namespace (on non-clusters) and host that
namespace on multiple servers. Some customers always use server clusters, so
they use stand-alone roots. Other reasons to use stand-alone roots are
greater number of recommended links (50K versus 5K for domain-based) and the
fact that you don't have to be a domain admin to create a stand-alone
namespace.

Now about the user data. Say you set up \\bigcluster\Mark as your root on
your cluster. Still, no user data is involved here...this is just the
starting point of your namespace. The server bigcluster runs the DFS service
and has some local file system structures to represent the namespace. Again,
no user data, just DFS internals. Next you would set up a series of folders
with targets in the namespace. like \\bigcluster\mark\secretfiles.
Secretfiles points to a UNC path, \\server1\CIA or wherever those files are
stored. A user who navigates to \\bigcluster\mark\secretfiles is directed to
\\server1\cia transparently. Well, what if \\server1\cia goes down? Then
users can't access those files UNLESS you make that target redundant. You
can do this two ways--store the user data on a cluster (thus giving the user
data redundancy) or replicate that data to another server, say
\\server2\cia. To the user, though, the path is just
\\bigcluster\mark\secretfiles, and assuming you've got multiple targets, the
user is directed to the closest server (in terms of AD site costs). The DFS
client can fail over to the other UNC path if one of the servers goes down.

I hope this helps! I highly recommend checking out the overview doc on
www.microsoft.com/dfs. This I hope will illustrate how these two
technologies work together.
--
This posting is provided "AS IS" with no warranties, and confers no rights.

Want to learn more about Windows file and storage technologies? Visit our
team blog at http://blogs.technet.com/filecab/default.aspx.
Post by MarkT
The request is for us to create a new DFS Namespace on a (2) node cluster.
The namespace would need to be a stand-alone DFS namespace because the
domain-based DFS namespace is not supporte on a cluster.
All the data would then be copied to the new DFS Namespace. Next, we would
build a second (2) node cluster and configure it as an additional server for
the namespace then use DFS Replication between the two nodes. But from what
your saying, the DFS Replication service is not cluster aware so this is not
a supported configuration.
So a standalone DFS Namespace is supported on a cluster but DFS Replication
is not. Is there any reason at all to put a DFS Namespace on a Windows
Cluster, as suggested in the DFS:FAQ
http://www.microsoft.com/windowsserver2003/techinfo/overview/dfsfaq.mspx
Post by Jill Zoeller [MSFT]
DFS Replication is the replication component, not the namespace component
that you refer to for stand-alone namespaces. What this means is that you
can't use DFSR to replicate content stored on shared cluster storage.
Can you clarify if you are also interested in replication using DFSR? I
can't quite tell if you are asking for DFSN (namespace) guidance or DFSR
(replication) guidance.
--
This posting is provided "AS IS" with no warranties, and confers no rights.
Want to learn more about Windows file and storage technologies? Visit our
team blog at http://blogs.technet.com/filecab/default.aspx.
Post by MarkT
We are wanting to implement Windows Clustering in two separate data centers
for the sole reason of hosting Windows 2003 R2 DFS. As I understand,
domain-based namespace is not supported on a cluster but a stand-alone
namespace is supported.
I am not clear on the following statement from The Step-by-Step Guide for
DFS
:(http://technet2.microsoft.com/WindowsServer/en/library/14211e1f-2dbd-418d-b721-2005c725ce351033.mspx?mfr=true
)
"On server clusters, replicated folders should be located in the local
storage of a node, because the DFS Replication service is not cluster aware
and the service will not fail over to another node."
1 - Can you clarify the above statement?
2 - Can you provide any resources for configuring DFS on a Windows Cluster?
3 - Other than the standard limits of DFS might you have any other
recommendations?
MarkT
2007-03-29 02:22:01 UTC
Permalink
Thank you very much !
Post by Jill Zoeller [MSFT]
Remember there are two things you want to have redundant here: (1) the
namespace and (2) the user data. These are both separate technologically.
They don't even have to exist on the same server.
The namespace has a service component that hands out referrals to clients.
If you want users to be able to access the namespace (i.e., receive
referrals), you need to make sure that the service can continue to run if
one of the namespace servers (aka root servers) goes down. One way to do
this is to host a stand-alone DFS namespace on a server cluster. Another way
is to create a domain-based namespace (on non-clusters) and host that
namespace on multiple servers. Some customers always use server clusters, so
they use stand-alone roots. Other reasons to use stand-alone roots are
greater number of recommended links (50K versus 5K for domain-based) and the
fact that you don't have to be a domain admin to create a stand-alone
namespace.
Now about the user data. Say you set up \\bigcluster\Mark as your root on
your cluster. Still, no user data is involved here...this is just the
starting point of your namespace. The server bigcluster runs the DFS service
and has some local file system structures to represent the namespace. Again,
no user data, just DFS internals. Next you would set up a series of folders
with targets in the namespace. like \\bigcluster\mark\secretfiles.
Secretfiles points to a UNC path, \\server1\CIA or wherever those files are
stored. A user who navigates to \\bigcluster\mark\secretfiles is directed to
\\server1\cia transparently. Well, what if \\server1\cia goes down? Then
users can't access those files UNLESS you make that target redundant. You
can do this two ways--store the user data on a cluster (thus giving the user
data redundancy) or replicate that data to another server, say
\\server2\cia. To the user, though, the path is just
\\bigcluster\mark\secretfiles, and assuming you've got multiple targets, the
user is directed to the closest server (in terms of AD site costs). The DFS
client can fail over to the other UNC path if one of the servers goes down.
I hope this helps! I highly recommend checking out the overview doc on
www.microsoft.com/dfs. This I hope will illustrate how these two
technologies work together.
--
This posting is provided "AS IS" with no warranties, and confers no rights.
Want to learn more about Windows file and storage technologies? Visit our
team blog at http://blogs.technet.com/filecab/default.aspx.
Post by MarkT
The request is for us to create a new DFS Namespace on a (2) node cluster.
The namespace would need to be a stand-alone DFS namespace because the
domain-based DFS namespace is not supporte on a cluster.
All the data would then be copied to the new DFS Namespace. Next, we would
build a second (2) node cluster and configure it as an additional server for
the namespace then use DFS Replication between the two nodes. But from what
your saying, the DFS Replication service is not cluster aware so this is not
a supported configuration.
So a standalone DFS Namespace is supported on a cluster but DFS Replication
is not. Is there any reason at all to put a DFS Namespace on a Windows
Cluster, as suggested in the DFS:FAQ
http://www.microsoft.com/windowsserver2003/techinfo/overview/dfsfaq.mspx
Post by Jill Zoeller [MSFT]
DFS Replication is the replication component, not the namespace component
that you refer to for stand-alone namespaces. What this means is that you
can't use DFSR to replicate content stored on shared cluster storage.
Can you clarify if you are also interested in replication using DFSR? I
can't quite tell if you are asking for DFSN (namespace) guidance or DFSR
(replication) guidance.
--
This posting is provided "AS IS" with no warranties, and confers no rights.
Want to learn more about Windows file and storage technologies? Visit our
team blog at http://blogs.technet.com/filecab/default.aspx.
Post by MarkT
We are wanting to implement Windows Clustering in two separate data centers
for the sole reason of hosting Windows 2003 R2 DFS. As I understand,
domain-based namespace is not supported on a cluster but a stand-alone
namespace is supported.
I am not clear on the following statement from The Step-by-Step Guide for
DFS
:(http://technet2.microsoft.com/WindowsServer/en/library/14211e1f-2dbd-418d-b721-2005c725ce351033.mspx?mfr=true
)
"On server clusters, replicated folders should be located in the local
storage of a node, because the DFS Replication service is not cluster aware
and the service will not fail over to another node."
1 - Can you clarify the above statement?
2 - Can you provide any resources for configuring DFS on a Windows Cluster?
3 - Other than the standard limits of DFS might you have any other
recommendations?
Sean Sullivan
2007-09-08 17:14:02 UTC
Permalink
Jill,

I know this is an old thread but it is exactly what I need help with so I
hope you don't mind a revival here. I've read everything I can get my hands
on and think I am about 90% there but need to validate a few things before I
set it all up.

My environment/requirement:
I have a Windows 2003 server cluster (2 node) atached to a SAN that I want
to use to consolidate my file servers into and add redundancy (current FS's
are stand-alone, local storage, no DFS).

I can set up additional clusters for additional components is necessary (is
it OK to host a stand-alone namespace and a file share resource on the same
cluster?)

I can handle the data replication using NetApp tools or 3rd party apps so
DFS Replication is not a requirement (I"ll gladly use it if it fits but if it
is not supported in the cluster then I can wrok around it).

My goal is 100% uptime at the primary site but also need to make data
available at the DR site and make it easily accessible (eg. I can get the
data replicated there without DFS Replication but would really want the
namespace to remain the same at the DR site so everyone could get at their
data easily).

So I think my dilemna is wheter to create a domain-based namespace that
contains links to clustered storage at the primary site and a file share at
the DR site (and I'd need to figure out the replication somehow) or, create a
stand-alone namespace on a clustered server that has links to clustered file
storage. But in case #2, how do I get the make the namespace exist at the DR
site?

In your example below, when creating the namespace on \\bigcluster\mark are
you creating a stand-alone namespace or a domain-based? Or does it matter?
The piece that I'm confused is the note about local file resources to
represent the namespace. Do these local resources exist on both nodes and the
cluster can handle them after failover?

If you were starting from scratch and could do it any way which is your
preferred design?

Thanks very much,
Sean
Post by Jill Zoeller [MSFT]
Remember there are two things you want to have redundant here: (1) the
namespace and (2) the user data. These are both separate technologically.
They don't even have to exist on the same server.
The namespace has a service component that hands out referrals to clients.
If you want users to be able to access the namespace (i.e., receive
referrals), you need to make sure that the service can continue to run if
one of the namespace servers (aka root servers) goes down. One way to do
this is to host a stand-alone DFS namespace on a server cluster. Another way
is to create a domain-based namespace (on non-clusters) and host that
namespace on multiple servers. Some customers always use server clusters, so
they use stand-alone roots. Other reasons to use stand-alone roots are
greater number of recommended links (50K versus 5K for domain-based) and the
fact that you don't have to be a domain admin to create a stand-alone
namespace.
Now about the user data. Say you set up \\bigcluster\Mark as your root on
your cluster. Still, no user data is involved here...this is just the
starting point of your namespace. The server bigcluster runs the DFS service
and has some local file system structures to represent the namespace. Again,
no user data, just DFS internals. Next you would set up a series of folders
with targets in the namespace. like \\bigcluster\mark\secretfiles.
Secretfiles points to a UNC path, \\server1\CIA or wherever those files are
stored. A user who navigates to \\bigcluster\mark\secretfiles is directed to
\\server1\cia transparently. Well, what if \\server1\cia goes down? Then
users can't access those files UNLESS you make that target redundant. You
can do this two ways--store the user data on a cluster (thus giving the user
data redundancy) or replicate that data to another server, say
\\server2\cia. To the user, though, the path is just
\\bigcluster\mark\secretfiles, and assuming you've got multiple targets, the
user is directed to the closest server (in terms of AD site costs). The DFS
client can fail over to the other UNC path if one of the servers goes down.
I hope this helps! I highly recommend checking out the overview doc on
www.microsoft.com/dfs. This I hope will illustrate how these two
technologies work together.
--
This posting is provided "AS IS" with no warranties, and confers no rights.
Want to learn more about Windows file and storage technologies? Visit our
team blog at http://blogs.technet.com/filecab/default.aspx.
Post by MarkT
The request is for us to create a new DFS Namespace on a (2) node cluster.
The namespace would need to be a stand-alone DFS namespace because the
domain-based DFS namespace is not supporte on a cluster.
All the data would then be copied to the new DFS Namespace. Next, we would
build a second (2) node cluster and configure it as an additional server for
the namespace then use DFS Replication between the two nodes. But from what
your saying, the DFS Replication service is not cluster aware so this is not
a supported configuration.
So a standalone DFS Namespace is supported on a cluster but DFS Replication
is not. Is there any reason at all to put a DFS Namespace on a Windows
Cluster, as suggested in the DFS:FAQ
http://www.microsoft.com/windowsserver2003/techinfo/overview/dfsfaq.mspx
Post by Jill Zoeller [MSFT]
DFS Replication is the replication component, not the namespace component
that you refer to for stand-alone namespaces. What this means is that you
can't use DFSR to replicate content stored on shared cluster storage.
Can you clarify if you are also interested in replication using DFSR? I
can't quite tell if you are asking for DFSN (namespace) guidance or DFSR
(replication) guidance.
--
This posting is provided "AS IS" with no warranties, and confers no rights.
Want to learn more about Windows file and storage technologies? Visit our
team blog at http://blogs.technet.com/filecab/default.aspx.
Post by MarkT
We are wanting to implement Windows Clustering in two separate data centers
for the sole reason of hosting Windows 2003 R2 DFS. As I understand,
domain-based namespace is not supported on a cluster but a stand-alone
namespace is supported.
I am not clear on the following statement from The Step-by-Step Guide for
DFS
:(http://technet2.microsoft.com/WindowsServer/en/library/14211e1f-2dbd-418d-b721-2005c725ce351033.mspx?mfr=true
)
"On server clusters, replicated folders should be located in the local
storage of a node, because the DFS Replication service is not cluster aware
and the service will not fail over to another node."
1 - Can you clarify the above statement?
2 - Can you provide any resources for configuring DFS on a Windows Cluster?
3 - Other than the standard limits of DFS might you have any other
recommendations?
Continue reading on narkive:
Loading...