SGI: hinv

As promised before , let me introduce my new workhorse:

hinv -mv


L2 controller


environmental data


serial data


pci settings

There are 4 cards spread out over the 4 buses:

• module1, bus1, slot1: IO9
• module1, bus2, slot1: LSI 22320 (dual U320 scsi)
• module1, bus2, slot2: Gigabit ethernet (modded HP NC7770)
• module2, bus1, slot1: LSI SAS3041X-R
• module2, bus2, slot1: Qlogic SANblade QL2342 (dual 2Gbit fibre channel)

Unlike most people I'm not planning to turn it into a wannabee-Onyx350 or Tezro. This one will live in the natural habitat of Origins: in a rack somewhere, running 24/7 and quietly performing it's duties (compile & compute server).


The second module is an expansion compute module, so it lacks IO9 and scsi drive backplane. It is still possible to install disk sleds, they just don't slot into anything. There is enough space for a SATA cable, though. You know where this is going

I briefly attached a 2TB Samsung to the SAS3041X-R and verified that IRIX can fx and mount it. I'm currently awaiting two extra disk sleds and two Samsung EcoGreen F4EG 2TB disks. Will make a nice backup server

Update 4-nov-2010 : added DVD-ROM and XVM stripe of two 2TB SATA disks.
Update 3-dec-2010 : increased RAM to 16GB, added QLA2342 dual 2Gb SANblade
Update 6-dec-2010 : added L1 data
Update 19-sep-2011 : IO9 firmware upgrade, added ethernet card
Update 15-nov-2011 : added L2 controller

Sweet!
Or... I know how sweet they are.

Congratulations - I know ramq hated to part with them, but they couldn't have gone to a better home.

Almost the same set up as my daily desktop <except that I couldn't resist the temptation to have a wannabee Onyx >.

Added 4TB of local storage: two 2TB SATA disks attached to an LSI SAS3041X-R, striped using XVM. These are "eco" (5400RPM) models. The SAS3041 is in the lowest PCI slot of the second Origin 350, where you would normally find the IO9. It has the PCI bus to itself so it runs at 100 or 133MHz PCI-X speed. I had to order some special 70cm SATA cables with an angled connector at the drive end: there's insufficient between the drives and the nearest fan for anything else. I expect some more RAM sometime soon, maybe I'll take some pictures then.

Disks installed:

Some results:

Compare this to the system disk, a Seagate 15K.3:

As was to be expected, the 5400 RPM disks loose to the 15K disk for small random I/O, but for everything else the SCSI disk takes a severe beating. Both tests use direct I/O, all disks have the (disk) write cache enabled. The system is in use, which probably explains the dip at block size 4194304 for the SCSI disk.

jan-jaap wrote: This one will live in the natural habitat of Origins: in a rack somewhere, running 24/7 and quietly performing it's duties (compile & compute server).

As somebody expecting delivery of two Origin 300s and a NUMAlink cable next week, I'm curious - were these received with licenses so they could be used as compile-servers? I'm new to linked nodes so I'm wondering about hostid changes and how one deals with licenses that no longer match... RTFM PMs and redirects to other threads welcome.

smj wrote: I'm new to linked nodes so I'm wondering about hostid changes and how one deals with licenses that no longer match... RTFM PMs and redirects to other threads welcome.

If they have previously been linked they'll already have the same system serial number. If they haven't previously been linked, the pair will assume the serial number of which ever becomes the master module <though you might have to use the "serial clear" command in the L1 of the *unlinked* slave so when it is linked it's ready to receive the serial number from the master>. So examining what licenses <if any> came with each module before selecting the master would be a good idea.

If you have an L2 controller <or emulator>, then the attached modules will assume the serial number assigned to the L2. A very nice fringe benefit of having an L2 controller is that can easily change the L2 serial number at will using the L2 command "serial set <desired_serial_number>". The 'serial set' command doesn't work at the compute module level <with the L1 controller>.

The following excerpts from an L2 "serial all" command example shows the serial numbers assigned to 01c001 <the master compute module>, 001c002 <a linked slave module>. Both also list the <same> L2 "reference system serial number". These particular compute modules didn't come from the same source and originally had different 'local system serial numbers'. One had a full set of MIPSPro licenses, the other didn't. The serial number from the licensed system was assigned to the L2, which in turn wrote the licensed serial number it to the <previously> unlicensed slave.
An L2 controller can be used to set the system serial number of any of the O3x0 modules *except* the Router Module, which has an enforced serial number security routine coded into NVRAM <that prevents end-user system serial number changes> - though there are now a couple of known workarounds for the NUMALink <Router> Module <and the R-Brick>. This would also be a good time to mention that you can't link more than two O3x0 compute modules without using a router - a NUMAlink Module <or a hacked R-Brick>.

jan-jaap: Sorry about the thread hijack -- I'll start one or more when my Origin 300's come in this week.

recondas: Thanks, I've started re-reading Escher's threads about his recently acquired O300's, and all the threads linked from those; grabbed the L3 software & DINA and put up a mirror... I'll see about distilling all of this into something for the Wiki, too, as it keeps coming up.

This system has genuine MIPSpro C/C++/F77/F90/Workshop and XVM licenses. Everything is 100% legit

Otherwise, the SGI FLEXlm licensing scheme hasn't changed since IRIX 6.2 / FLEXlm v4.x. I won't get into the details because it probably violates your awful DMCA law and I don't want to cause any trouble, but it isn't hard to break.

Some pictures of the SATA mod:




Needed:
* Two O350 drive sleds or the Intel / SUN equivalent
* Two 70cm SATA cables with a straight connector on one end and an angled connector on the other end
* Two SATA disks. I used 2TB Samsung disks
* A SATA power cable with two vertically attached SATA power plugs and a Molex
* Another misc power cable with at least a Molex plug
* One TYCO ELECTRONICS / AMP - 3-640426-4 - HOUSING, 18AWG, 4WAY

Of course you cannot boot from this, and the drive bay must not have the SCSI drive backplane either. So this works only in a system with at least two modules.

Good looking installation - thanks for sharing the photos and parts list!

Great idea! Since I took the IO9 out of my second O350 this would be a perfect upgrade for that brick.

Space between the disks and the blower is limited. I looked around for a suitable SATA backplane but found nothing. Actually, I suspect that the Altix 350 has it. If someone is going to gut one, feel free to donate it to me But that's why the cables have to be just right. I had the right SATA power cable in my junk bin, but not the extra long, angled SATA cables. They have to be the 'left' / 'straight' model :
There are some clips on the outside of the O350 chassis, and the space to route the cable around the fan, which is why I chose the left hand cable. The more common cables with vertical attachment won't work on the lower drive (not enough space).

I bought a big box of 2GB memory kits so this system now has 16GB installed. Unfortunately 1 DIMM is bad or dirty so the machine currently runs with 2GB disabled and I don't have the time to pull it from the rack and rectify that. As soon as that's fixed I'll update the 'hinv' one (last ?) time. That's also the reason I put the SANblade in: if the need to attach some SAN storage pops up I don't want to have to pull the system from the rack ...

After a bit of cleaning all 16GB RAM is now enabled

Updated L1 to rev. 1.48.1 (IRIX 6.5.30 + patchSG0007149) using the procedure described by tjsgifan (thanks!)

This changed a few things in the '* env' output: both warning RPM and current RPM of the PSU fan have mysteriously doubled on both modules. Also, the '* serial all' now probes a SNOWBALL (nor found) and serial numbers of the PSUs. Speaking of those, I get this now too:


Right

The whole procedure is rather involved, if it wasn't for the fact that I was rebuilding an O300 anyway I wouldn't have bothered.

smj wrote: Good looking installation - thanks for sharing the photos and parts list!

indeed. very nice machines and just as nice work. good to see so many mips cuties still in action

Found an affordable 147GB Seagate 15K.5 to replace the 73GB 15K.3 system disk:

The dip in bwd_wt at 512K/1024K is reproducible, but it's still a nice improvement

Updated the IO9 firmware.

We have seen a fair number of ethernet adapters fail here, so I'm now using a modded OEM (HP/Compaq) 'tg' card, rather than waiting for my IO9 to suffer a mysterious death.

Added L2 controller. It came without power adapter, but thanks to this post by Recondas , that problem was quickly solved.

I'm using THIS power adapter. I had to order the DD50 connector in (multiples of) qty 10, and the backshell in qty 5. If someone wants to repeat this exercise, you can have a connector + backshell for 5€ + shipping.

Was greeted with this a couple of days ago:

One of the SATA disks gave up after running 24/7 for 6 years straight (this system had not been rebooted in over a year).

I had a spare disk, but this was a stripe so now I'm reloading 3.5TB of data from backups ...

I hope this is not ruining your weekend, Jan-Jaap?