8

I have an original ML370, from before there even were generations (2000). The storage system is comprised of a Smart Array 431 controller, 2 9.1 Gb Ultra3 SCSI (COMPAQ BD009635C3), and 4 146.8 Gb Ultra3 SCSI (COMPAQ BD14687B52).

The two 9 gigs form array A which forms one RAID 1+0 logical which is the OS drive. The four 146.8 gigs form array B which is carved up into about 12 RAID 5 logicals. Regrettably, some of those logicals are mounted by way of "junctions" over system directories; "inetpub", "Program Files", "Database" (SQL Server), and "Message Store" (Exchange)... imagine this were Linux and that would make perfect sense, although on Windows that's questionable.

Also, for what it's worth, the network adapter is a Linksys EG1032.

At this point everything (almost) has been migrated off of this server, however, there is still one legacy application that can't be migrated due to bus architecture and driver support for current operating systems, nor is it in line to be replaced at this time. So, it's necessary to keep this box running for a little while longer.

Enter storage problems; one each of the 9 and 146 gig drives is failed, so the entire storage system is degraded and needs to be addressed ASAP. So the actual question is, what are my options?.. and BTW, the intention is to not spend another dime on this thing.

This server is running Windows 2003. I'm concerned that it may not be possible to reinstall everything from scratch since so much time has passed. I have an iSCSI SAN at my disposal and there would be sufficient space on the remaining three 146 gig drives to run two as a mirror and keep the third around for a spare. The challenge seems to be how to go about reorganizing the existing storage while keeping the server running while doing it. The 431 isn't very smart.

So I'm wondering, can a server of this age be made to do a network boot? My first though was to get it running off of an alternate logical disk of some type so that I could tear the existing arrays apart and reconfigure them. But I'm open to alternate suggestions. I know what I can do with modern servers, but something of its age I know is much more limited and I'm not sure of the options.

TIA

ewwhite
  • 194,921
  • 91
  • 434
  • 799
tlum
  • 257
  • 1
  • 8
  • 1
    15 years is... a pretty good run. Its probably old enough if things started failing, things like ram and such would be hard to replace. I'd also suspect considering how much more efficient modern systems are that a modern replacement would pay for itself in a few years. Also. 8gig disks. My phone has more storage... – Journeyman Geek Apr 21 '15 at 04:06
  • Windows Server 2003 end of support : 14 july 2015. – Guillaume Apr 21 '15 at 08:37

2 Answers2

15

I totally don't want to discount the rest of your question, but the path of least resistance is to replace the failed disks.

When dealing with legacy equipment, there may be a temptation to devise an esoteric or clever workaround, but your immediate goal is to keep this particular hardware running in the current and stable state until you can migrate everything off of it. Change as little as possible.

This is an original Compaq Proliant ML370 (white). I built and deployed a handful of these systems early in my career, and definitely supported them through the mid-2000's...

The important thing to know is that the original (G1), G2, G3 and G4 ProLiant servers all used variants of parallel SCSI (Ultra-2, Ultra-3 née Ultra-160 and Ultra-320) with 80-pin SCA connectors to the drive cage backplanes. This technology was supplanted by SAS and SATA.

Despite this, SCSI disks are still available. The Smart Array 431 RAID controller is intelligent enough to accept a replacement drive, so there's nothing to worry about there. You can buy the original products (9GB and 146GB) on eBay quite easily for a few dollars.

Do that.

Meanwhile, take the necessary steps to move your legacy application off of the 15 year-old hardware. If that configuration runs the business, then it's a serious operational risk and the business leaders should be prepared to deal with the fallout.

ewwhite
  • 194,921
  • 91
  • 434
  • 799
  • The problem is there is some very expensive hardware interfaced to this box which means purchasing a modern $10k replacement in order to retire the ML370. All of the software applications were migrated years ago, what remains is a hardware compatibility problem. I'm also concerned with the fact that all of these drives are the same age... two down, four to go who knows how soon. The remaining storage requirement is a fraction of what it once was, so it's advantageous to scale back the number of disks that are on-line, but that's a lot easier said than done. – tlum Apr 21 '15 at 08:37
  • 2
    Just buy some used like-model drives. They're not manufactured anymore, but you can pick up a few disks (and extras) at a very, very low cost. You'll never be able to predict when or how disks fail and I doubt that the others are just going to die now. Again, why are you adding complexity to the situation? It is not in your best interest to make changes to this fragile setup until you have a viable path to replacement. – ewwhite Apr 21 '15 at 09:12
  • I hear what you're saying, however, it's been failing and I'm trying to just stay ahead of it. I hadn't mentioned the power supply had just died and I managed to fit another one I had laying around. The one drive was bad, and before I was able to get the space on that array consolidated, the one in the BOS array failed. When stuff gets this old it's not a matter of if it will fail, but how soon. Reducing the hardware to the bare minimum reduces the failure risk exposure. On the one hand you can keep throwing hardware at it, but that has been a chore and there appears to be no end in sight. – tlum Apr 21 '15 at 17:52
  • 1
    Trying to do anything other than hair-on-fire replacing it is absolutely going to be an exercise in throwing hardware at it. So, Ed's right. Fix the failed hardware; while doing so, keep yelling at the business that this whole server needs to be replaced before you can't fix it anymore. – mfinni Apr 21 '15 at 20:09
  • 1
    @tlum I'm sorry. It seems like you want to take a more complex approach to resolving this issue than is really necessary. If you want to netboot the machine, go ahead... If you think it's safe to move disks around or attempt to reconfigure a Smart Array 431 controller, you're risking downtime, data loss and angry business users. Hard drives, fans and power supplies are consumables, and the parts are clearly available for mere dollars. You didn't mention the PSU failure earlier, but you should either doing the bare-minimum to keep things running (_my suggestion_) and/or a forklift replacement. – ewwhite Apr 21 '15 at 20:31
-1

  1. some iSCSI implementations will allow you to export virtual drives onto SAS or SCSI (with appropriate hardware). and given that the 2003 box has an external SCSI connector this may not be difficult.

  2. the hardware will netboot without too much trouble (it supports pxe), but unless you are already netbooting 2003/XP boxes setting up the infrastructure to do so is a pita.

  3. you can add a pci esata card and use it with an external esata raid.

  4. you can use a sata raid card or sas card and use newer drives.
hildred
  • 216
  • 1
  • 3
  • 8
  • I have a local CentOS repo and net boot/net install VM guests now, so I'm equipped to pxe boot but Linux is the only thing I'm using it for right now. I wasn't sure if I could get that server with the NIC to net boot. I have to look into that some more if it seems plausible. – tlum Apr 21 '15 at 08:45