Douglas A. Tutty wrote: > Hello again. > > In my search for low-MHz machines, at least on eBay, I find lots of old > Compaq Proliants (all around the $300 mark by the way). E.g: > > 4500R: P-133, 1 GB ram, no drives, $249. > > HP doesn't have on their website the owner's manuals for these old > boxes, but they do have the service manuals. > > Googling around doesn't tell me anything about using OpenBSD on them. I > don't know right now which e.g. smart array they use. > > Is anyone aware of any Gotchas for these old boxes?
I've warned you about a lot of them, you ignored that, but for some reason I feel obligated to try one more time. I just hate to see people do things like this to themselves (and I want to be able to say, "No, not interested in helping on this" in clear conscience). For that kinda money, they better be delivering it...and helping you get it on the rack. Old Compaqs are an art. Old Compaq servers are a black art. They are some of the quirkiest, strangest, and most obnoxious systems I've worked with. Kinda like a Cisco switch, in that once you get the dang thing running the way you want, you feel so great because the pain stopped, so you tend to forget it just shouldn't have been that way. I've yet to see a multi-Pentium and only one Multi-PPro machine run OpenBSD/SMP. (score is at least two Pentiums and two PPros that didn't work with SMP), I suspect our EISA support has suffered severe bit rot. Between the system and the bus, I'd be rather surprised if you got the thing running OpenBSD (pleasantly surprised, yes, but surprised). If you do, please post dmesg. :) I just looked through the dmesg log, I saw no Pentium class EISA machines that people sent dmesgs from. I saw a few PPro systems, one PPro running GENERIC.MP, several Alphas and HPPA systems. The CMOS battery is dead (or will be soon). It isn't going to be easy to replace. See the SPARC Battery FAQ and the part about cutting into the old CMOS chip to solder in your own battery (it works, done it on a SS2 and a mvme88k, worked. I also seem to have toasted another mvme88k doing the same thing, but I didn't pay $300 for that machine. BTW: I'm way out of practice, but I'm still much better than your average $5 soldering iron novice, I used to do component-level repair on computers and other such things. I got good equipment and I sorta know what I'm doing...and I still managed to break the CPU board on the mvme88k. EISA isn't fun when it works properly. I've probably config'd more EISA machines than most people on this list, trust me, it's not fun. If you have never done it before, the time to learn was back in the 1980s, not now. WITH THE RIGHT TOOLS, Compaqs were some of the easiest to configure, but finding the right tools was exciting last time I tried. When it DOESN'T work properly...ew. No disks...you better hope they include the Compaq config utilities on a CD so you can install 'em and configure the thing. I've done it from floppies, Not Fun. I screwed up the disk config, reinstalled. More Not Fun. I think I did this three or four times. I did learn disk OpenBSD disk configuration Really Well, so I guess it was a good, not fun thing. Hope they include disk trays. There are a lot of old servers laying around, there are a lot of old disk trays. The servers and disk trays are rarely in the same place. No idea how that happens. There are several variations of Compaq disk trays, not sure how cross compatible they are. (68 pin drives, 80 pin (SCA) drives, 1" drives, 1.6" drives). Did I mention that Compaqs config the disk array using the utility partition or the utility CD? I have a stack of cac(4) cards. Spent a day or so building an array on a Windows machine, moved it to the target machine, and then discovered that cac(4)s are really, really slow. BTW: don't think that because you use SCSI, you don't have to worry about disk size. Expecting to build a 1TB disk array on a 15 year old controller may expose some "issues." <speculation> Old cac's have some kind of battery on them, they look like large lithium cells. They don't really look like rechargeable. Even if they are, they are so old, they are probably dead on yours (and mine). That may be why my cac(4) experience was so uninspiring, or it may just mean the things will toast your file systems when the power goes out unexpectedly. </speculation> Read, memorize, live by: http://www.openbsd.org/faq/faq4.html#cpq16m "some" most likely includes you. Three words: "Power Hungry Pig". Bringing up an old workstation is usually just a matter of taking your PC skills and remembering a little history. Bringing up an old server is a test of patience. Bringing up an old Compaq is a little like restoring a rusty old car, 'cept none of the neighbors walk by and say, "wow, that's cool". One of my favorite dis-features of the Compaq servers of around that era was a case interlock: remove the cover and the thing would save itself from...what, eventually overheating and maybe crashing?...by immediately turning itself off. (I discovered this disfeature at a client's site -- I was sent in to do a server upgrade. Never seen the system before, had zero trust that they had sent me with the right parts. So, about 15 minutes before scheduled "get out of the system" time, I started gently sliding the cover off the machine, so that if something was clearly not right, we could cancel with minimal impact on the users. I missed the big, bright blue(?) warning sticker. "Click" and silence, followed by me saying, "oh shit", thinking of all those people working frantically to get their work done before the shutdown time...) HOWEVER, in your case, you might actually prefer that it does shut itself off if you break the RF shielding. :) If you really want to go that route, look at what the machines are SELLING for on e-bay, not what people are asking for them. BIG difference. Offer $50 to the person asking $300 for the machine, you will probably get a "no way in hell", and the machine a couple weeks later. :) Nick.
