The collected works of smj - Page 8

You're at least right to the extent that somebody clueless might notice it in the logs, and there'd be another round of "Chinese haxxors" in the news...

I neglected to ask previously -- was your colleague's machine in a working state, and then stopped? Or was it left unused for a few years, and you're now trying to revive it? Might be a factor - I'm not sure how the Indigo2 should behave with a flat/dead NVRAM battery, at least one person got garbage video output rather than a blank screen. But I know a few workstation types will fail early in the boot process without a live NVRAM.


Good point, and easier to test than soldering up a serial cable. (Again it's worth checking eBay - €14 delivered? Then again, it may take some time to arrive...)

You want to remove all SIMMs from the machine, then populate the four slots closest to the rear of the machine (marked "RAM A" or "BANK A" on the system board) with four SIMMs of identical size and speed. Give it another try; if it still doesn't make it to a yellow power LED, a chime or video output, you could try swapping in four different matched SIMMs, if you had more than four to start with.

Hmm, how much (more) of my idiocy do I want to expose here... The two bays in the 1U server are for the mirrored system disks. That only leaves me the four drives to work with - which are in a ZFS raidz1 with weekly scrubbing, in hopes that scrubbing might warn of approaching danger. Going with raidz2 /RAID6 would in this case give up 50% of raw capacity, which is harsh... Not as harsh as failure, but that's always easier to accept afterwards.

Just checked the "Power_On_Hours" and they've actually only been running about 1.8 years - but they are Seagate's of a troubled 2009-2010 vintage/model (ST31500341AS). Some folks have had so much trouble with those drives they'd run screaming right there, but I think there's some luck involved and the reported issues are exacerbated by bad interactions with hardware RAID controllers vs. my 9212 HBA. That said, when I check "Reallocated_Sector_Ct" I see values of [2, 53 , 1, 0]. Hmm - and smartd(1m) was mis-configured until recently, so I don't know if that 53 grew slowly over time, or has recently entered a death spiral... <_<

Well, guess I'd better plan for that transition - the SE3016 would have enough bays to get redundant. I'm unlikely to be looking at anything spinning faster than 7.2k, so I doubt raw bandwidth will be the limiting factor even at 3Gb with 8 drives.

Wow - that is pretty neat! But I've got one squirreled away somewhere, and at $230 (I have the other bits) I'm not too tempted...

I knew there should have been another large chip on that CPU module, but I had no idea on the VSIMM - what gave it away, all the chips I didn't notice were missing on one side of the PCB?

Factory spec'd sure, but for maximum pain entertainment I wonder if the "up to 33MHz" SM20 would function in an Aurora...

From the car hobby I gather there's a 100 pound limit on Greyhound Freight nowadays. Guys were discussing sending their gearboxes that way, 94 pounds plus packaging was going to be close...

Hey, no problem! We're all enthusiasts and happy to help. Don't be a stranger, and don't forget to check out the NekoWiki for lots of useful info - especially the Indigo2 page , I just added a bunch of stuff there the other day.

Short answer -- if you're already adding another, you might as well surrender now and get both, you're clearly doomed.

What are your general criteria for choosing to collect a particular model? What I mean is, did you pick up the I2 because of it's significance in the evolution of 3D workstations? Because it looks cool? Because of the software it ran/runs?

If you really intend to max out either an Octane or Fuel, it's gonna add up fast...

The Octane is definitely an iconic SGI workstation, from an industrial design or packaging perspective. While it has the dual CPU option and up to 8GB of RAM, you're far more likely to find those CPUs running from 225-400MHz than 600 and something like 2GB of RAM. Meanwhile the Fuel's CPU usually runs at 500-600, though yes up to 900 in rare cases; it likely has the same 2GB of RAM instead of the 4GB max, but it is DDR; and the SCSI bus is outright twice as fast. If you want PCI cards, the slots are right there - no need to find shoehorns or a shoebox. Fuel has USB built-in, but you'll have to hunt a little for an audio card. Octanes are loud, but fan replacement's supposed to make a big difference.

Without fan replacement an Octane isn't going to be doing regular duty in my office, I don't think. Fuel's a much better choice there. But the Octane is just a whole lot cooler in general, so...

Just curious hamei, I don't think I've seen this yet and I was listening to some "Best of 2012" coverage of the IPO -- what do you make of Mark Zuckerberg?

Y'all are making me nostalgic for my 3130...

(Which had an ESDI controller and a pair of Hitachi DK51_ drives - ~170MB each? I forget, it was reasonable and at the time (~1990) it was even conceivable to get some branch new 300 or 600MB ESDI drives to upgrade. Sigh. mex ftw!)

Embrace the madness, crrn!

Audio card: Get the model number off of it. Photos are fine, but you can just lookup the board in the aggregator or on the Fuel NekoWiki page . There are only a few that are supported, but random no-name USB audio adapters seem to work very well (again, see aggregator or just go here: viewtopic.php?f=3&t=16725440&p=7341670 )

DO NOT tell him about the innate coolness of single system image IRIX running spread over several different physical boxes. Next thing you know he'll be looking for CrayLink/NUMAlink cables and asking questions about IR/g-bricks...

Ya, good idea, that's a section of the wiki I haven't looked at yet. And something FAQ-like would be good, new or expand what's there. I've run myself ragged this past month tuning up the car club's web site, but hopefully soon...

There, fixed that for you.

So no combination of two of your four DIMMs inserted in slots 0 & 2 will allow it to pass POST, but all four slots full will? That's odd. Myself, I'd retry the combinations and make doubly certain that for each DIMM insertion the card-edge fingers really go in the socket all the way - I've had that problem before with these critters... By "make sure" I mean grab a flashlight and stick your head in the machine, and if need be use an inspection mirror.

Also, I suspect this is where having a terminal attached to the motherboard serial port might be helpful as you power up.

Clusters are usually collections of standalone machines, each with their own OS instance running, and use various services or libraries to coordinate execution of independent executables over some or all of the machines in the cluster. They often use some kind of library to mask the fact that cluster-wide memory is broken out in chunks across the machines and cannot be addressed uniformly by any single process.

When you start plugging the later SGI servers together, you're converting standalone machines into one shared memory multiprocessor. In other words all those boxes click together like Legos (literally the code name of the SGI Origin 2000) to make one big computer, running one instance of IRIX.* If you have 5 Origin 300s each with 4 CPUs and 4GB of RAM and 2 73GB HDD, cable them together through a NUMAlink router and you wind up with a single computer that has 20 CPUs, 20GB RAM, and 10 HDD.

When you run IRIX on that collection it just looks like one large machine to your software - a Single System Image (SSI). The app you were just running on your Indigo2 can suddenly schedule threads / child processes on 20 CPUs, mmap() ~20GB of RAM, etc without rewriting to use other libraries, configuring MPI, hooking up GigE or Infiniband, etc. And at the multi-gigabyte per second speeds of the NUMAlink connections.

I realize 20 cores and 20GB of RAM would fit comfortably in a tower case under your desk for the past year or three, but you can't as far as I know put a second box there, plug a cable between the two, and have a shared bus machine with 40 cores, etc. But you could do this stuff with the Origin 2k (and $$$) back in 1996...


* You can "partition" such a system, putting a subset of the system's resources into an apparently-independent machine. Similar to Solaris "zones" I believe, and similar features from other systems.

I will not call you old fashioned, but rather wise. Only now when I check with Word 2010 do text styles seem to work properly - probably only because in two minutes I can only make a trivial test...

Because I'm not familiar enough with the latest Intel E-series Xeons and Bulldozer Opterons to know whether or not you can get more than 16-24 cores into a tower case that'll fit under a typical desk. It made sense in context.

Interesting, I had a 3/60 w/ a P4 color framebuffer on my desk ~1990-91, but just one mono 3/60 in storage. These both appear to be CG4s (501-1210 and 501-1248), to Pentium's point - slightly different implementations, but functionally the same.

I'm really curious about the 4 pin Molex connectors on the one board, wondering what that was for... Nice to see the old 386i style shoebox on top of the pile.

I'm out of my depths here, but my reaction is A) don't bend the boards while powered; B) if the board he's bending is fitted to a backplane slot, it isn't seating well, try cleaning the fingers (if relevant) and re-seating; C) if it was more like an Indy or Indigo2 mainboard, I'd want to carefully examine the solder joints of the through-hole pins on any connectors like power, backplane, etc. The close to the point of flexing, the more closely I'd examine them. But I humbly defer to somebody with relevant training in these matters.

I get where you're coming from, certainly relative to a whole keg-like Octane, but - you have looked at the PIMM in an Octane, right? They tend to have a huge metal heatsink and PCB carrier attached. Not that it's a Herculean task to pack one, but it does take a bit more attention to detail and materials. It certainly isn't as trivial as pulling a Xeon out from under it's heatsink, wiping off the thermal grease, popping it in an anti-static bag and into padded mailer with $1.50 of stamps on it...

Ah - that must help explain why Storage Services requires us to buy each disk we want them to use to provide storage to our machines at EMC list prices, then charges us a metered fee each month based on how much of that space we're using, in addition to the fees just to have the machine connected to their SAN several times for multipathing and all that. And thankfully I've forgotten the details of how they rape our budgets for tape-based backups - or used to, maybe they've stopped by now (was still relevant ~4 years back).

Meh. Let's go back to speculating about Illumos/OpenIndiana/OpenSolaris...

Also I'm pretty sure the slots at the back of the case are Bank A, versus B and C. Check TechPubs but I'd start with the lowest ordinal bank and work up from there. I've got 1GB in an R10k I2, using the HP D NNNN parts mentioned on the NekoWiki I2 page...

Ian's page is the reference I used, so not sure what I can add to that.

Hmm - thanks for posting that snippet, apparently I've been filling the SIMM slots backwards...

Really?


Kudos!

I get spreadsheets all the time where the only reason they used Excel was because they found it easier to lay out tables there than in Word. (Assuming they knew it was possible to create tables in Word...)

This practice ranks right up there with the one where they publish an article on the super expensive, cutting edge, best-of-breed content management system that runs the internal web portal, where the article says "Something important, please read >this<." where "this" is a link. To a Word document, stored on a different content management system like FileNet/P8 or Sharepoint. And in the Word document - after everything finishes chasing links and starting several more apps on your PC - are two paragraphs of plain, unformatted text...

In light of the HSBC sanction-subverting and money laundering settlement, where the government basically declared them not "too big to fail," but "too big to prosecute" -- I don't know if Enron is a contrast to changes post-2008 crash, or if Ken Lay was simply guilty of thinking too small...

I was going to suggest UNIXPackages.com , but they only have packages for SPARC and x86. Might want to dig around on Bill Bradford's site SunHelp.org and see if you can find something there. If not, and nobody here has a better suggestion, ask on one of the mailing lists at SunHelp.

Forgive me if I'm being stupid -- is there in fact a how-to somewhere? I may have missed a link...

Forgive a basic question, as I'm not familiar with IBM's line -- did IBM release a BIOS update or firmware patch that definitely supports Harpertown/5400 Xeons in this model? There are plenty of servers (ex. HP DL140 G3, Dell PWS690) that have a chipset capable of supporting them, but the system won't because the vendor didn't implement support in the BIOS. The DL140 uses the Greencreek 5000X chipset, which can support Harpertown according to Intel, and while the 140 will correctly ID them at power-on if installed it cannot proceed to BIOS or boot. (I might be confusing which model behaved in which way, but neither would boot an OS.)

Don't get me wrong, they're fine people, good Americans.
But they're content to sit back, maybe watch a little
Mork & Mindy on channel 57.
Maybe kick back a cool Coors 16 ouncer.
But they don't know what the queers are Android is doing to the soil.

Dead Milkmen FTW!! Beelzebubba was a tour de force. Mods, feel free to move this to the 7 page "Music" thread.

I was listening to something - a "tech" podcast I think - about how the iPhone and anything with Apple's logo prominently featured had been the darling of the fashionistas in Hong Kong and Korea for quite a while, but that the bloom was off that rose - and that as they went, much of Asia would follow. They asked somebody in Thailand about it who said, "Yeah, we pretty much do whatever Korea does."

Another point one panelist was making was that Apple is only available on something like the sixth largest mobile carrier in China - and therefore when they did complete the pending deal with China Mobile (?) there were another ~750MM potential iPhone buyers. So hamei did they release the iPhone with US-style lock-in there, or did this guy reveal a lack of knowledge about how the China mobile market works?

Heh - you guys should invite Tony Duell from classiccmp to chime in, but I'm not sure how well nekochan works when you're using an ASCII-based web browser...

Hadn't heard about that, but then I've somehow missed all of Amazon's cloud stuff so far. However I had been thinking something like S3 might be an option for off-site backups after the data's been suitably encrypted. Probably already been incorporated into Amanda or Baccula by now. Hell, do it using that filesystem layer they built on top of GMail back when a free 5GB mailbox was astounding.

John, do you mean the VS4000 console in "glass TTY" mode, e.g. no window system running? To be honest I'm not sure it does support anything like decent terminal features.

If it is in console mode, did you have trouble installing DECwindows, starting it, or did you opt-out during the installation? I think it's pretty straightforward to install DECwindows after the fact, get the media mounted and run one or two DCL scripts, but I'd have to hit Google...

I'm afraid I'm posting mostly to offer moral support - I don't recall an issue with the Apple CD-ROM drives, but I was probably using them with DECstations rather than VAXstations. I was just gifted with a VAXstation 3100/m76 (gave a bunch away, but this came with a NeXTstation), but I'm not sure how quickly I can get to trying to set it up.

One question, which media are you using? By any chance, a CD-R that you burned? Just prompted to ask by the reminder about CD-R compatibility in commodorejohn's VS4000/m60 thread... I realize it's booting, but you never know.

Heh - you're doing fine, thanks. Confirmed what I thought might be the case, that the guy was assuming too many parallels between mobile service/companies in North America and China.

They didn't say it, but I assumed HK might set the trend for the mainland, while Korea would influence the rest of the Rim -- with a weird interference pattern sitting over Japan such that one wave from Korea they embrace, the next they skip - rinse, repeat. But I'm probably just revealing my own unfounded speculation there.

Thanks again!