Suppose you have an external storage system, raid5, connected via SCSI to a very old HP Prolian 5000 (1997). Suppose the Proliant goes down due to Power Supply failure.

Suppose, waiting that the part gets replaced, that somebody decides to disconnect the external storage system, connect it to another proliant server (3500) in the naive hope it will recognize the raid5 chain; boot the proliant 3500, find out from Windows NT 4.0 disk manager that the disks are there individually, and not as a raid 5 chain, switch off the server, and reconnect to the Proliant 5000.

Now, I am back to work and find out what happened. Tomorrow I will get the part, and reboot finally Proliant 5000 . Will the raid 5 chain be there, with all data intact, or should shoot myself right away? In other words, are there chances that the raid5 configuration is lost, and if so, how can i recover it?

Having done time with old Compaq hardware... Assuming that you have a Smart Array Controller inside your disk array, the controller is what would configure the RAID5. As long as nothing was done to touch the disks (formatting, adding drive labels, etc.), you should be okay.

Retrieve the RAID controller from the dead machine and have it ready to install in the new machine. That method should have a very high probability of working.

Now as to what NT does when it suddenly finds itself running on completely new and unknown (to it) hardware, well..

Cheers

Sometimes NT asks the question about writing some kind of a signature to a new disk, claiming it is harmless. Anyone know if THAT would harm his RAIDs?

I had this exact same thing happen. We installed dummy hard drives into the RAID controller for the first boot. Then shut down and intalled the drives with data on them.

As Paul said, assuming it is an external unit that was connected to a Smart Array, simply connecting it to another Smart Array should bring things back. 1997 era storage products that were Compaq branded were just dumb SCSI disk holders, with no array intelligence inside them. Thus it was likely cabled to a Smart Array PCI card. Based on the age of the boxes, I'm guessing a 2/P or 2/SL.

Writing a disk signature to each individual disk would likely cause damage, including possibly complete data loss. I can't remember exactly where on the disk the NT signature goes to say for sure though.

One world of advice. Try and make sure the Smart Array controller you plug these disks into has at least the same, if not newer firmware then the controller they came from. That should help in the auto migration. If you want to check that everything worked right, make sure you have a SmartStart CD handy to boot to. That has the Array Configuration Utility on it, and will be a better test to see if things are back, before letting an OS poke at the disks. For servers of that era, SmartStart 5.5 is the last version they will boot, and it is free to download.

I actually never bothered to find out exactly what it was that NT writes to the disk when it asks that quesiton. That's something I should probably know.

/me researches.

http://www.microsoft.com/technet/archive/winntas/support/diskover.mspx

"The disk signature is a unique number at offset 0x1B8 that Windows NT uses as an index to store and retrieve information about the disk in the Registry subkey HKEY_LOCAL_MACHINE \SYSTEM \DISK."

From other mentions in that article, I get the impression that it's four bytes at or near the end of the master boot record. So the next question is... Does tweaking four bytes in the master boot record ruin the RAID array?

Also, the question is if it got that far. I always thought that you had to agree to add the header, and that only really happened in Disk Manager. If you just powered it on, it might see the drives (POST test), but it shouldn't do damage unless it prompted and was accepted.

Unfortunately the person realized the disks were there, indivudually, from NT Disk manager. So he actually got that far.

Oh... That's not good...

If the drives did have the signature writen. R-Studio should recover the information but you will need a place to image all of the drives in your array. I used it to recover files on a friends mirror array after he had used a ghosting type program to change his partition size on the mirror.

http://www.r-tt.com/FAQ.shtml#q13

THis is major help, Attack. Thank you. I'll try to have that sw ready in case of major disaster.

Good news. Replicated what happened in our lab, and the Smart Array didn't even blink. Here is what I did:

1. Ran around the entire lab for a while just trying to find an NT4 box still working
2. Found a Proliant 7000 with 2 hard disks in a RAID 1 on a Smart Array 3100ES controller
3. Pulled the disks out and moved them to the NT 4 Proliant 6500 with no Smart Array.
4. Opened Disk Administrator and said yes to writing the signatures.
5. Powered down the NT 4 box and moved the disks to the Proliant 7000
6. Booted the 7000, watched as it saw the disks and all the data on them

So it looks like the Smart Array ignores the MBR and only pays attention to the config data stored in the first 4-8mb of the disk.

Drakino, THANK YOU!
What can I say, this board, meaning all of you guys, is fantastic!

No need to say, the part is not here yet.
I spent the day in meetings discussing the political reasons why are we in such a situation: who was were, why happened what happened, and, to make a long story short, it's not a pleasant time here. And all this is caused, ultimately, by the fact that the department did not want to spend money in updatin the machines when they should have, or renewing the maintenance contracts on that old old box. Of course, when it crashed, they put pressure on a person who knew little on how to solve the problem.

Well, Drakino, your test results are comforting, very very much! Really, thank you!

Please note, these power supplies tend to go out in pairs. If you had multiple power supplies and replace one in a Compaq Proliant 5000, be ready to replace others. We just had a bunch of them fail here in some of our older machines. Just a word of warning.

Wow!

Paul, thank you for the advice, totally appreciated. But you won't believe this: that machine has ONE power supply. Besides being 8 years old. And besides having no maintenance contract on it. This is to give you an idea, if you should ever need, of how narrow minded can organizations be. Oh, and I don't want to forget to underline how long ago I had requested funds to replace it: THREE YEARS. And I've been asking over and over. If I wasn't stressed enough for the most embarassingly obvious and predictable disaster that could haver happen, I would be happy it finally happened.

And the most ridiculous thing of all is that the new server is located two rack units above it, and migration was in progress when the old one crashed. No, I'm not joking. Really, I shouldn't, but I feel like laughing

And I don't want to mention now why our backup is not very useful in this particular case, because I am still too mad.

Really, guys, thank you again. This place is a breath of fresh air, expecially in days like today.

Did you ever get that power supply? How did this end?

Yes. Yesterday. And the new power supply was... broken! (broken connectors, precisely).

So, "Tomorrow" (how many times I heard this?) we'll get another, hopefully working, power supply.

I am going crazy after this. Really. I have no more words to express how frustrated I am.

Thank you for asking. I will definitely post here when I reboot the server. If I don't shoot myself before I get the part, of course.

... so, today's news is that the part is coming... tomorrow

This is endless.

Yesterday the Power Supply arrived. Wrong model. But same specs, so we connected it and... Data was there!

Thank you again Drakino for testing, and keeping our hopes alive! And thans everybody for your help!