MPIO on win 2003 and shared SAN volumes

[baggies1]

So i have two 2003 servers.

One is connected to a SAN volume and both have the MultiPath IO software from the vendor installed.

My questions is this. I want to connect both servers to the same volume.

Whilst MPIO is designed to ensure no corruption occurrs on the SAN for two paths from the same server does it still work for two different servers connecting to the same SAN volume?

 

[p5wizard]

Short answer: NO.

Long answer:
MPIO is a means to provide multiple (or alternate) paths to a disk so that a (ONE) server does not loses access to its disks when one path fails. Mostly it is used for SAN LUNs and the paths are through different HBAs and different fabrics and different SAN server controllers, but a SCSI or SAS disk array may also have two connections that may be connected to different SCSI/SAS adapters in a server to provide an alternate path if one cable gets pulled or if one adapter fails or if one connection in the SAS array fails.

To avoid corruption in a disk shared between more than one server, you need some sort of clustering software that can keep track of which server is accessing which (part of a) disk.

HTH,

p5wizard

 

[RMGBELGIUM]

You need indeed a clustered filesystem to make this work.

NetApp Certified NCDA/NCIE-SAN

 

[baggies1]

Thanks guys.

So as i understand it i need "windows server" to do the actual clustering.(the san in the background is on a diskarray).

One of the two servers i want to connect ot the same volume is only server 2003 Standard.

Can you recommend a third party tool to do this?

I'd read IBM SANergy can be used for this? (and is free to download?)

 

[p5wizard]

> So as I understand it I need "windows server" to do the actual clustering.(the san in the background is on a diskarray).
Yes.

> Can you recommend a third party tool to do this?
No, sorry I can't...

I'd read IBM SANergy can be used for this?
Perhaps, but I have no experience with this software

> (and is free to download?)
I doubt it...

HTH,

p5wizard

 

[baggies1]

ok thanks all.

I dont really have clustering setup experience and i have inherited a system which hasnt been designed greatly with redundancy in mind. Our primary DC also hosts an old mail system (icewarp)

There are 3 dc's (2003) , two of which are "enterprise" allowing me to cluster (right?).

Once i've run the cluster admin setup and added the other "enterprise" dc node, can i then safely reintroduce the SAN share to that DC without fear of corrupting data?

For some reason i've got this funny idea about cluster being a replication of resourses across both servers but i think this only happens if you actually go to both servers and install the resourses on both manually. The cluster sits at the front end. Clustering doesnt actually replicate those resourses but just sits at fron end.
Is that a good assumption/interpretation?

 

[kmcferrin]

You're probably better off in a Windows forum for this, but this is the way that Windows Cluster works:

Lets assume that you have Server1, Server2, and SAN1. On SAN1 is a disk volume named LUN1.

Server1 and Server2 are both connected to the SAN. LUN1 is presented to both servers. In order for clustering to work both servers need to be running Enterprise edition of Windows. Additionally, MS only supports clustering on identical hardware. But assuming the clustering requirements are met, LUN1 is presented to both servers but only one server will be able to see it at a time. If you look in Disk Management on the second server, you can see that LUN1 is there but it will show as offline. If the node that can see LUN1 is shut down then the other node will be able to see it.

When you configure clustering, you will configure resources that are clustered. Those resources can be a cluster name, IP address, disk, file share, print spooler, application, script, etc. The resources are configured to be available on both servers, but are only owned by one server at a time. In the event that one of the servers goes down, the other server takes over processing of resources.

It's not clear what you're trying to cluster for high availability, but it sounds like you might be trying to cluster AD or domain controllers. That won't work, and in fact a domain controller shouldn't be used as a failover cluster node.

It's actually fairly complicated setting up clusters in 2003, so if you don't know what you're doing then I highly recommend that you do some reading first so that you understand how they work.

You can start by watching this webcast by Symon Perriman from the MS Clustering and HA team:

https://msevents.microsoft.com/CUI/...lture=en-US&EventID=1032364830&CountryCode=US

________________________________________
CompTIA A+, Network+, Server+, Security+
MCTS:Windows 7
MCSE:Security 2003
MCITP:Server Administrator
MCITP:Enterprise Administrator
MCITP:Virtualization Administrator 2008 R2
Certified Quest vWorkspace Administrator