iSCSI and 2003 Cluster

All,

I have a 2 node Server 03, SP1 (don't ask) cluster that is currently using
dual single port HBAs for connectivity to a Clariion CX3-20c SAN. I
connected one NIC from each node of the cluster to a switch dedicated to
iSCSI traffic and configured iSCSI on the cluster with v2.06 of the initiator
software. I can see the LUNs from the servers but I'm just curious more than
anything if someone has some real world experience in making iSCSI work on a
2003 cluster? I did find this article,
http://exchangeexchange.com/blogs/jo...-clusters.aspx
and it seems pretty straight forward. I'm going to post this question in
the storage and 2003 server groups as well but I just wanted some thoughts or
any known "gotchas" that someone may have run in to before.

TIA,
Dean

That is fantastic! Let me work through your instructions and I will let you
know. Thanks for the great detail!

Dean

"Plain Simple" wrote:

> Hi Dean,
>
> Just setup my 2003 server two node cluster with iscsi (using Dell
> Equallogic SAN), what is it that you have questions on?
>
> Here are the basics:
>
> To Do for both nodes
> 1. Install Windows 2003 Server SP1 (or whatever is newer)
> 2. Install Microsoft hotfix to allow support for GPT disks (disks
> above 2tb) http://support.microsoft.com/kb/919117/
> 3. Install Microsoft iSCSI initiator
>
> Take inactive node offline
>
> On the primary active node
> 1. Connect to iscsi targets and LUN's
> 2. Partition and format LUN's according to your desire, also assign
> drive letter(s)
> 3. Create a new group in Cluster Administration named "iscsi" (or
> whatever you want)
> 4. Add a resource to the group for a "Generic Service", the service
> name is "MSiSCSI"
> 5. Take the group and resource online
> 6. Bring the second cluster node online and fail over the "iscsi"
> group to it
>
> On the second node with "iscsi" group now active on it
> 1. Connect to iscsi targets and LUN's
> 2. Assign drive letter(s) to appropriate drives
> 3. Add another resource to the "iscsi" group for a "Physical Disk",
> point it to the partition you want to be clustered
> 4. Bring the "Physical Disk" resource(s) online
> 5. If the resources come online successfully, fail the "iscsi" group
> over to the primary node
>
>
> Done!
> .
>

Sorry, one other thing. I want to continue using my FC as my primary
connections to the SAN, I just wanted the iSCSI as a backup in case of a FC
disaster..like loosing both switches, as improbable as that is. So all of
the data is already formatted and I see the LUNs via FC. Would I simply
continue with step 3 of your instructions?

Thanks,
Dean

"Plain Simple" wrote:

> Hi Dean,
>
> Just setup my 2003 server two node cluster with iscsi (using Dell
> Equallogic SAN), what is it that you have questions on?
>
> Here are the basics:
>
> To Do for both nodes
> 1. Install Windows 2003 Server SP1 (or whatever is newer)
> 2. Install Microsoft hotfix to allow support for GPT disks (disks
> above 2tb) http://support.microsoft.com/kb/919117/
> 3. Install Microsoft iSCSI initiator
>
> Take inactive node offline
>
> On the primary active node
> 1. Connect to iscsi targets and LUN's
> 2. Partition and format LUN's according to your desire, also assign
> drive letter(s)
> 3. Create a new group in Cluster Administration named "iscsi" (or
> whatever you want)
> 4. Add a resource to the group for a "Generic Service", the service
> name is "MSiSCSI"
> 5. Take the group and resource online
> 6. Bring the second cluster node online and fail over the "iscsi"
> group to it
>
> On the second node with "iscsi" group now active on it
> 1. Connect to iscsi targets and LUN's
> 2. Assign drive letter(s) to appropriate drives
> 3. Add another resource to the "iscsi" group for a "Physical Disk",
> point it to the partition you want to be clustered
> 4. Bring the "Physical Disk" resource(s) online
> 5. If the resources come online successfully, fail the "iscsi" group
> over to the primary node
>
>
> Done!
> .
>

"Dean" <> wrote in message
news:37B60484-150A-4CBE-9F34-...
> Sorry, one other thing. I want to continue using my FC as my primary
> connections to the SAN, I just wanted the iSCSI as a backup in case of a
> FC
> disaster..like loosing both switches, as improbable as that is. So all of
> the data is already formatted and I see the LUNs via FC. Would I simply
> continue with step 3 of your instructions?
....>
just to make sure what your goal here is. If you want to use ISCSI as a
backup storage scenario be sure, that this works not without a storage based
block level replication here. There are several vendors who can do
replication of 3rd party storage systems but I'm not sure if they also cover
ISCSI targets -> Recoverpoint (EMC) or Safeguard (Unisys). Otherwise it will
not work on a automated way like usual in cluster.

With manual intervention (restore data from backup to ISCSI, change
ressource in cluster groups....) possibly you can work out a desaster
recover scenario....but not recommended :-(

If I did understand your question wrong please let me know.

Regards
Ramazan

Ramazan,

Hopefully I can clarify this. What I want is simply to have iSCSI
installed and configured so if my fibre channel connection dies, my cluster
continues to access to my SAN. I don't want to use is as the primary
storage. Can this be done on a cluster or I guess any server for that matter?

Thanks,
Dean

"RCan" wrote:

>
> "Dean" <> wrote in message
> news:37B60484-150A-4CBE-9F34-...
> > Sorry, one other thing. I want to continue using my FC as my primary
> > connections to the SAN, I just wanted the iSCSI as a backup in case of a
> > FC
> > disaster..like loosing both switches, as improbable as that is. So all of
> > the data is already formatted and I see the LUNs via FC. Would I simply
> > continue with step 3 of your instructions?
> ....>
> just to make sure what your goal here is. If you want to use ISCSI as a
> backup storage scenario be sure, that this works not without a storage based
> block level replication here. There are several vendors who can do
> replication of 3rd party storage systems but I'm not sure if they also cover
> ISCSI targets -> Recoverpoint (EMC) or Safeguard (Unisys). Otherwise it will
> not work on a automated way like usual in cluster.
>
> With manual intervention (restore data from backup to ISCSI, change
> ressource in cluster groups....) possibly you can work out a desaster
> recover scenario....but not recommended :-(
>
> If I did understand your question wrong please let me know.
>
> Regards
> Ramazan
>
> .
>

This is Windows 2003, and to be supported you need to have your full cluster
on the Microsoft HCL, including hosts, firmware, HBA's, switches, and
storage.

So, you need to ensure that your vendor has a tested and qualified cluster
with iSCSI and FC on the HCL
Else, this will not be supported

rgds,
Edwin.



"Dean" <> wrote in message
news:37B60484-150A-4CBE-9F34-...
> Sorry, one other thing. I want to continue using my FC as my primary
> connections to the SAN, I just wanted the iSCSI as a backup in case of a
> FC
> disaster..like loosing both switches, as improbable as that is. So all of
> the data is already formatted and I see the LUNs via FC. Would I simply
> continue with step 3 of your instructions?
>
> Thanks,
> Dean
>
> "Plain Simple" wrote:
>
>> Hi Dean,
>>
>> Just setup my 2003 server two node cluster with iscsi (using Dell
>> Equallogic SAN), what is it that you have questions on?
>>
>> Here are the basics:
>>
>> To Do for both nodes
>> 1. Install Windows 2003 Server SP1 (or whatever is newer)
>> 2. Install Microsoft hotfix to allow support for GPT disks (disks
>> above 2tb) http://support.microsoft.com/kb/919117/
>> 3. Install Microsoft iSCSI initiator
>>
>> Take inactive node offline
>>
>> On the primary active node
>> 1. Connect to iscsi targets and LUN's
>> 2. Partition and format LUN's according to your desire, also assign
>> drive letter(s)
>> 3. Create a new group in Cluster Administration named "iscsi" (or
>> whatever you want)
>> 4. Add a resource to the group for a "Generic Service", the service
>> name is "MSiSCSI"
>> 5. Take the group and resource online
>> 6. Bring the second cluster node online and fail over the "iscsi"
>> group to it
>>
>> On the second node with "iscsi" group now active on it
>> 1. Connect to iscsi targets and LUN's
>> 2. Assign drive letter(s) to appropriate drives
>> 3. Add another resource to the "iscsi" group for a "Physical Disk",
>> point it to the partition you want to be clustered
>> 4. Bring the "Physical Disk" resource(s) online
>> 5. If the resources come online successfully, fail the "iscsi" group
>> over to the primary node
>>
>>
>> Done!
>> .
>>

"Dean" <> wrote in message
news:5BB68145-3814-4FB6-ADF3->
> Hopefully I can clarify this. What I want is simply to have iSCSI
> installed and configured so if my fibre channel connection dies, my
> cluster
> continues to access to my SAN. I don't want to use is as the primary
> storage. Can this be done on a cluster or I guess any server for that
> matter?
Dean,

If I got you right you want to have an sort of MPIO between an ISCSI and
Fibre connection ? If so, the answer will be no - there is nothing available
which can handle this at that time (or nothing I'm aware) :-(

Even if your SAN is capable to provide ISCSI and FC to your server the
active paths cannot be handled by any MPIO framework which I'm aware now.
And if there are not the same SAN boxes how did you handle the data
integrity (realtime syncron replication) ?? MSCS application (a.e. SQL) does
not like inconsistencies ;-)

Can somebody confirm ?

Regards
Ramazan