*Sonic*

Got a question on SAN's

Just put in an MSA1000 Small Business, with Small Business High Availability Upgrade kit

2 servers connected, 1 is a Windows 2003 server, the other a Windows 2000 server

Both are accessing the same 1TB array, but not at the same time, the main reason is incase the main 2003 server goes offline for any reason

A couple of shares are on there using DFS, no replication, and the idea is that if the main server goes down, then new dfs links can be added to the backup server which is windows 2000 and connected to the SAN

I seem to be having a problem tho, in that sometimes one of the servers will report that some of the folders are unreadable

If I reboot one of the servers it goes into chkdsk mode

Is this a problem, as one is Windows 2003 and the other Windows 2000, using slightly different versions of ntfs (the Array was created under Windows 2003, and formatted using Windows 2003)

they both have the MPIO drivers loaded, which provides basic failover, however I do appear to getting a large number of inconsistencies with either file i/o or unreadable folders

the only seemingly way around this is to shut down both servers, and power on the Windows 2003 server first, and skip the chkdsk stage, once up and running, then power on the Windows 2000 server, and again skip the chkdsk stage

Any help would be appreciated

Used SAN Surfer Express to configure it all

Steve

KiwiGTI

I didn't think you could access SAN volumes concurrently.

HHxx

Err that is a really odd way of doing it..

I would use the same os on both and run it using cluster services. Win2k Advanced or Win2k3 enterprise iirc for cluster services.

This way you can have virtual servers setup in cluster services and resources like shares, ip, name.... The cluster groups can then be hosted by either physical node and failover etc... DFS can then point at the virtual server names.

btw, 2k3 and 2k ntfs are slightly different...

I'm not quite sure how it all works in your current setup as I have never done it that way before for failover resilence.

H

*Sonic*

The customer didnt and doesnt want clustering, despite me trying to talk them into it, I do know all about clustering tho

the windows 2000 server is going to be upgraded to windows 2003 over the next week or so

The original spec was both to be run on windows 2000 OS, but as the management software doesnt run on windows 2000, we had to make an emergency alteration and make the 2003 server part of the san, so it could be managed and installed and configured

what happened this afternoon, when i let chkdsk run was that it decided to re-index all the security descriptors after a good hour or so it had only done 120,000 files out of the 798,000 so i came home leaving it running

My guess that it did this due to setting permisions from the win2k server, which had to be done this way

I admit it is a strange setup, and told them today that they need to upgrade to 2003 asap

HHxx

I think I'm catching your drift

The problem I see is nothing is controlling who owns the disks...

Sounds like you need a nas solution

*Sonic*

LOL they didnt want a NAS solution either !!

I know what you mean about who is controlling the disks, they both will want to write a signature to the disks

They really only wanted it in case the server controlling the SAN failed, and with a few changes to dfs paths, a quick reboot, and all would be working

all the hardware is redundant etc, and clustering would be a better solution, but they didnt want the additional cost on top of what they had already spent

KiwiGTI

Then i think you are pretty much stuffed, the design just won't work. You will be in an endless cycle of Chkdsks etc.

Where's the additional cost in clustering? You've obviously got the h/w already which would be the biggest expense.

*Sonic*

Would windows 2003 on both servers alleviate the running of chkdsk each time a server reboots?

They dont really want clustering, and they really only want to bring the other server online if the main one fails

I guess one option could be to leave the fibre disconnected on the backup sevrer, and only plug them in if the other server fails, they dont need 24/7 protection as such


biggest expense would be having to buy two copies of Windows 2003 Enterprise edition, then the expense of installing the OS (upgrade etc) then setting up clustering

KiwiGTI

No, it's not really the fault of the O/S. There needs to be something that controls the disks which has to be clustering, NAS etc.

SAN's are not like Windows shares where you can have multiple machines accessing and managing the same volume.

The disk signature also maps directly to the servers registry so any other server will need a different disk signature to be written.

*Sonic*

hmm, maybe i need to force them to use clustering, not really an avenue I wanted to go down