The collected works of smj - Page 16

Jack's original comments and J-J's follow-up seemed to indicate that it's normal for these systems to dynamically assign IDs to the drives. Seems pathological to me but as others have indicated, I've avoided Fibre Channel and SANs for the most part.

A close examination of the drive PCBs may reveal jumpers somewhere on the underside, if you don't find them on one end or the other. Take the model number and try to find the spec sheet on the web, that should tell you where to find them. Then get some jumpers and fix the addresses - that should stop the madness of them popping up in different places every power cycle, if I understand what's happening.

Good luck! All this tells me is, stay the heck away from these "stones" - they'll drag you under!

ISTR installations on an Octane with R10k CPUs would not work when moved to an Octane with R12k CPUs. Not sure I tried the reverse. (That was probably using 6.5.19 FWIW.)

I think that was before I got my hands on any VPro gfx cards - where you will also face a problem swapping over from MGRAS. There are threads you can find on nekochan about adding the necessary software to switch gfx card families,

Send me your dual R14k CPU module and I promise I'll run all the permutations and report back...

robespierre: What you say makes sense, thanks. Looks like my original impulse to resist opening my mouth was correct - I don't know about this, and am not helping...

I'll tell you something though - I believe I've got Storage Enclosure Services in my SE3016 JBOD. If it ever starts changing what IDs the drives get without my having moved them around, it's getting replaced...

Or you could setup DINA ...

A 900MHz PIMM? Outstanding! More RAM is always nice, but 1GB actually goes a long way under IRIX unless you're doing Serious Work.

Unfortunately finding a V12 may be time consuming and/or expensive.

Alright, y'all made me pull out some papers and check... I'm in Berkeley, California and service is from PG&E. I have 17 solar panels on the roof, so I "just" pay for whatever I pull off the grid (500-900 kWh/mo), but it may not be clear how it translates to any standard residential rates.

Anyway looking at April-February billing starting in 2014, I averaged $0.247/kWh with a range of 3.6 cents (22.8 cents to 26.4 cents). I usually use $0.25/kWh for estimating year-long operating costs for electric lights and such, so nice to see I wasn't so far off.

Two replies before yours, by tomvos - the tarball's right there...

I haven't seen a nekochan-specific one, but is there any reason that Bountysource couldn't be used?

The three ISOs from SLES 9 SP3 for IA64 were a little under 2GB on disk. If it's the right ProPack, "ProPack4_for_Linux_SP3" is about 600MB in three ISOs.

I'm not sure *what* is in that 34GB blob, but I'll find out tomorrow...

When dealing with an SGI machine that has an L1, the magic keystrokes to switch between L1 and console mode are ^D and ^T - that is, CONTROL-D and CONTROL-T. I forget which is which, but the system will tell you which one to use to switch _back_ when you enter the appropriate one.

Note that when you switch from console to L1, by default it's just for one command, then returns to console mode - enter "l1" (I think) to stay in L1 after switching from console mode.

Applications have to be specifically written with the intention of using MPI at runtime. It doesn't just distribute any old program across a given set of machines.

I had never heard that Maya was built that way, but I'm not a Maya user so I might easily have missed something. What makes you think Maya was built to be used with MPI?

tooR wrote: In 2003 my mac xserver g5 and workstation g5 quad xgrid show one machine 15Ghz for any app.

Sure, but did Xgrid let you decide one day to run Photoshop, or Microsoft Word, or SimCity across those machines without modification? Only if the author/company creating the app designed in Xgrid support, would something like that work. Or if you built your own app to use Xgrid. Or perhaps if you had something that involved lots of small command-line jobs, like compilation or batch rendering, where you could start the command line jobs via Xgrid's command line tools.

Generally speaking, you need something designed to run in a distributed manner. Even the most common examples that aren't, like building code with distcc - the compiler isn't specifically designed to be distributed, but the compilation controller ("distcc") definitely is.

Maya can probably spread back-end rendering jobs across multiple machines - again, I'm not a Maya user, but it seems likely. But MPI won't let you automatically parallelize and distribute the front-end GUI application at runtime.

I thought it was Challenge/Onyx (1) PSUs, and Altix/Origin 3x0 PSUs, that were reported to burst into flames - not Octane PSUs...

I'll defer to Adrenaline on the experimental part - the A350 is stowed, and I need to focus on the new job. But SLES 11 SP1 has been rock solid on the A350 and A450 examples I've got here.

foetz wrote: now that's a stock start

Fixed that for you.

That's right, that's absolutely right. Because in fact, hamei's never seen my house. He's seen my fake house, which has some furniture in it, because my real house is packed wall-to-wall with Twiggy-drive Lisas and Macs, Cray-1s, dozens of max config IRIS 3130s, Xerox Altos, original Connection Machines, the fabled prototype "Minnow" desktop PDP10 connected to a Symbolics Framethrower, unbuilt Altair 8800 and Heathkit H11 kits, NeXT prototypes based on the Motorola 88000, a complete set of every SGI desktop machine with clear plastic skins, a PDP-11 running OS/2 that I can't tell anybody on classiccmp about, and I have Gary Kildall and Steve Jobs - both deaths were faked - locked up in my basement.

And they're all in or on awesome purple SGI racks.

That's right chumps - y'all got nuthin' on me.

If any of them can make a decent tavë kosi or çomlek , and the truck is loaded with rakia , keep 'em coming!

Remember, Alver was an HP minion once upon a time -- he could've actually gotten hold of the ZX6k deskside kit. I read (somewhere) that the deskside config was a lot quieter...

recondas wrote: Might not be a bad idea to turn the bounty concept around the other way. In order to request a software build/port, you have to test at least three beta apps and post your findings.

Now you're cookin' with gas!

uunix wrote: If I'm wrong I will eat humble pie and stand in the corner.

Waiting for pictures...

I'm glad that the problem you were seeing seems to be due to the networking, which you can now correct.

But I couldn't help being confused by this:

ivelegacy wrote: I’m rather surprised about that: my O2+ with R12K @ 400Mhz is faster than my friend's impact R10K @ 195Mhz
the feeling you have sitting in front of the machine is that O2+ is faster than impact
and Mathematica for Irix completes things faster

Well, yes - your O2+ *is* faster than his Indigo². Short of serious graphics tasks where IMPACT might come into play, why do you think a machine with a CPU half as fast should perform better? Even with Mathematica, all the math is happening on the main CPU, not in the gfx system.

If you're looking for a general rule - avoid the 1U formfactor if you want to escape the noise. Anything limited to ~40mm fans has to have a couple fistfuls of them, and spin them fast, in order to keep things cool even under partial load. Moving to 2U can make a big difference as the 60-80mm fans aren't as noisy, and you have more options for replacing those sizes with quieter versions.

That doesn't mean the Apple x86 Xserve's can't be quiet - I haven't seen/heard one in person. Like I said, just a general rule...

hamei wrote: I have an HP switch that is the exception to that rule : they took a 45* chop out of the bottom and laid a big fan over at an angle. Pretty clever way to get some quietness and a lot of air movement into a 1U case.

I can't conveniently check now, but I'm pretty sure you'll find the same thing in a NeXTstation slab...

Interesting news! However I will immediately delete this from my brain, because:

tjsgifan's suggestion sounds promising, especially if you haven't already checked all the connections inside.

It's probably easier to wait for your NUMAlink cable and poke at the problem brick from the working brick. But... You could grab the L3 software ( Use the Search, Luke ! - there are multiple forum threads that cover it) and use it to see if you can talk to the problem brick over its L1 USB connection from a PC or VM.

I understand and sympathize with the motivation behind this - I really do. Just... I wouldn't get your hopes up.

For the past efforts: By 2003 SGI was committed to Altix/Prism/VW xx 0, and Windows or Linux, as the core product set. Ignoring Windows, I can understand how it would make sense to extract things that were unencumbered and could be useful to differentiate these platforms/systems. And part of the marketing around Linux was participating broadly in FOSS, so opening up the parts you were working on actively made some sense. XFS is a great example because that gave them a useful differentiator and selling point, but they weren't cutting their own throats by exposing most of it, since their competitors already had alternatives. As for 4Dwm and other workstation bits, they were committed to Windows on the one hand, and I guess they decided they didn't want to deal with having to support it on Linux on the other, so it didn't happen.

These days Rackable/SGI does nothing with IRIX, and they're not about to start doing anything with IRIX again. The question at this point is, what motivation do they have to reassign busy engineers and pay them to go back, relearn the code base, separate out the interesting and unencumbered bits that they didn't already pull out in previous efforts, package each component or subsystem up, explain to the lawyers why they should give it away for free, convince more lawyers they won't get sued by somebody looking to make a quick buck, and then actually release the stuff?

I'm not saying don't do it - heck, I signed one or two petitions around this in the past. Just don't be too shocked if the response from Rackable/SGI is ... underwhelming.

I doubt it will attract much (if any) more interest, but FWIW that sled design originated in legacy Compaq storage systems. They were most common after the acquisition of DEC, but I believe their were a Compaq design even though they were often marketed under the StorageWorks brand post-acq. I believe you'll find them in late model Compaq x86 servers lke the DL360 G3, storage systems like the StorageWorks 4100 RAID array, as well as later Alpha-based servers like the DS20/ES40.

Ah trivia, when can I have those neurons back?

I've enjoyed all of the half dozen or so books of Pratchett's that I've read so far. Which is no mean feat, given that they were all pretty much selected at random. Well, Colour of Magic was probably recommended, but Hogfather , Going Postal and Making Money , etc were picked up based on his track record. Sad news, even if expected.

You should really add your location to your profile...

wenp wrote: I'd say OS X is the least Unixy Unix-like system out there. ... no /proc

If /proc is a canonical feature of UNIX, somebody has performed a major piece of retcon on the space-time continuum...

Axatax_ wrote: I'm now an "official" MPlayer dev

Woo hoo! Congrats!

And you are located, generally speaking, where? A country, perhaps? A continent, at least?

I picked up (via Craigslist) the corner unit guardian452 shows, plus a matching table that goes well next to it. Came with a keyboard arm fitted, I added cable races all along the backs of both surfaces, and a twin monitor stand out of frame on the edge closest to the camera.

Obviously getting it used was cheaper, and unfortunately you can't count on anything like that. But, it's an excellent setup.

Poor quality photo, that's a few years out of date now... You can tell because there's nothing piled up on the floor.

R-ten-K wrote: Huh? this project does not use the GPL, or has anything to do with it. Why bring it up?

Why not bring it up? When you're reviving 12 year old threads, the gloves are off - anything goes!

vishnu wrote: Do I spy with my little eye... A real Henry Miller Aeron?

Yeah, that was the one good thing I got out of my former employer for taking away my office and forcing me into the "work from home" program. And to the best of my knowledge that office still stands vacant about three years later - they haven't even reassigned my phone number.

vishnu wrote: Supposedly there are aerons aplenty in the surplus market but it hasn't seemed to drop their prices appreciabely, you're the only person I know who's got one, are they as good as is claimed?

No question there are excellent chairs out there for much less than what an Aeron will set you back, but I find them to be quite cromulent. Very comfortable, and nicely configurable if you get all the bells and whistles. Just make sure you don't get the wrong size. I think they do three, A - B - C, or small - medium - large. I had a large and I was a medium, and the ergonomics were off enough to be irritating -- but that's surely true of any chair...

So I wouldn't recommend everybody rush out and buy one or anything, but yes they're very nice. I had a given allotment that had to be spent through the employer's purchasing system, so since I didn't really need anything besides a new chair... Wound up with the Aeron and a Brother B&W multi-function widget, since I couldn't have the company laptop speaking to the house networks while the VPN was up.


To the original point, building work surfaces out of solid wooden doors is also a good option. Get yourself some file cabinets for a base, and you're good to go.

Note that I always disable SElinux so that it doesn't screw up all manner of things on me in non-obvious ways...


In late 2010, I built a VM using VMware Fusion on a MacBook, and installed Fedora 9/i386. Post-install (but pre-L3 installation) the kernel was updated to 2.6.27-25-78.2.56 - not sure if that would happen today, or what kernel version it started with. Sadly, I couldn't find a copy of the VM when I looked for it a few months ago.

When the Fedora 9 packages were all updated post-install, the mouse was screwed up. VMware KB artcile w/ fix here .

So, do you have an "sgil1" initscript in /etc/init.d, and does it load correctly? I have a fragment/note that says


I moved the VM over to a Thinkpad T23 running Fedora 13 or 14 and VMware Player before actually running it (successfully) against a two brick O300.

I got it to work under VMware - doesn't mean it can't be done under VirtualBox. In my case it was Fedora in side the VM and Fedora hosting the VM, when the L2 emulator was connected to the O300 via USB.

cf. viewtopic.php?f=3&t=16719038&start=15#p7286984 and subsequent posts...

xiri wrote: Hi,
I can confirm that it works well under CentOS 5.11 using VMware Workstation or VMware player.

Thanks xiri. Just to clarify - what's running in the VM, and what host OS did you use? And if you then connected the VM to the O300 via USB, any special steps required to allow the VM to access the USB port?

I'm with skywriter on this one. However loopy Stallman's philosophy, the packages that were developed and distributed early on under the FSF/GNU umbrella were a godsend. BSD UNIX source code was difficult to come by off-campus prior to the 1990s, and if you did so surreptitiously there was always a lingering fear of AT&T's lawyers... Life with UNIX before the Net/2 release, and really before the resolution of the USL vs. BSDi lawsuit in '94, was very different.