After working pretty much non-stop the last couple of months I decided to take a few days off ... well, everything ... but the internet in particular. Vanessa is visiting her family for Christmas. I stayed home with the cats, cooked some great food, drank lots of tea and read a bunch of books:
- In the Beginning Was the Command Line by Neal Stephenson. I've read this before but stumbled upon it going through my books.
- Across the Top of the World by James P. Delgado. About finding the Northwest Passage.
- Thinking with Type by Ellen Lupton. Typography handbook.
- Longitude by Dava Sobel. Descending from several generations of watchmakers I found this hugely interesting. It's about solving the navigational problem of finding the longitude by way of chronometers. And an interesting story of technological pragmatism and craftmanship vs. theroretical science. I bought this book a while ago and never found time to read it. Recommended!
- The Emacs Lisp Reference Manual. I first read this when Emacs and I started dating in the early 90s. I am one of those people that uses Emacs for pretty much everything but I'm mostly a copy-and-paste Lisp programmer so it was good to revisit this book and tinker a bit with Lisp outside the scope of .emacs hacking.
- The New New Thing by Michael Lewis. I inhaled this in a single reading session. It's about Jim Clark who started SGI, Netscape, etc. And about the .com bubble in general. I borrowed this book from Master Wilcox a while back and thought I'd better read it and give it back. Put both my time at SGI and our woes at Linuxcare in a whole new light. Especially so given that our interactions with Kleiner Perkins happened after the events described in the book.
In my lab I have a Dell PowerEdge T300 that's been my main test machine for a while. It was bought with SATA and the smallest drive possible because I always netboot my test boxes and don't need local storage.
A few months ago I got a Nehalem box that I've increasingly used for my testing. And consequently I decided to make the T300 my build box.
My old build box had a bunch of nice SAS drives that I intended to move to the T300. However, the T300 didn't have a suitable SAS wiring harness. Also, I decided to leave one of the SAS drives in the old build box. So I needed an extra SAS controller.
Off I went to Dell's website. I ordered a SAS controller for the T300. And then I spent a couple of hours trying to find the part number for the matching SAS wiring harness.
My T300 was wired for SATA so I needed a SAS data cable (SAS is dual-ported and has slightly different connectors). I needed SFF-8484 to 4 x SFF-8482. Also, in the T300 power and data are wired into the same connector so the wiring harness also had to provide a 2x5 pin hookup to the power supply.
I looked and looked. Turns out the harness is only listed on the US support site, not the Canadian one (The part # is NP390 in case anybody is interested).
I called up Dell's parts department, gave them the part number and two days later I had the cable in the mail. Yay!
Today the SAS controller arrived. And guess what? Inside the box was a wiring harnesses suitable for my T300. Serves me right for trying to outsmart the system. When you order a controller for a T300 you actually get the right cable. D'oh! I don't think it's worth the hassle returning a $17 part, though. So now I have a spare...
Sun, 07 Dec 2008
A quick guide to glitch-free audio on Fedora 10
The release note advertises glitch-free PulseAudio in Fedora 10. What the note conveniently forgets to mention is that the trick in getting glitch-free audio is to completely remove PulseAudio from the system. When PA is involved, playback stutters like Porky Pig on the North Pole...
This comes as no real surprise as removing PulseAudio was also imperative in the previous Fedora release unless you were completely tonedeaf. It consistently played back audio almost a semitone sharp on two of my machines. Both were using the extremely rare (dare one say "exotic"?) intel i810 audio hardware.
I applaud the tremendous progress we have made in making Linux a world-class desktop operating system.
Wed, 03 Dec 2008While the code has been out there for a while, Oracle formally announced the contribution of my data integrity enhancements to Linux yesterday. There's also a podcast and Emulex' website features a video interview with our fearless leader.
Thu, 30 Oct 2008
Check it out! It's groovelicious!
Wed, 15 Oct 2008
My slides from Open Source Days 2008 in Copenhagen are available here.
Tue, 30 Sep 2008I'm back home after a couple of weeks on the road. First stop was presenting at the Linux Plumbers Conference in Portland, OR. Great city, great conference. My slides are here.
After the LPC I flew to San Jose to present at the SNIA Storage Developer Conference. I believe SNIA will eventually make the slides available to the general public.
While I was away my DSL modem at the lab blew up. It's the second modem that's gone bad within a couple of months. Something really fishy must be going on with the phone wiring in that building. Most annoying.
Wed, 10 Sep 2008
Fri, 15 Aug 2008
Wed, 13 Aug 2008
In my two test JBOD arrays I have a whole whack of different drives. udev gives us the possibility to access disks by UUID, id, path or filesystem label. Most people use the latter. However, my test drives generally don't have filesystems on them. And certainly not filesystems with a lifetime long enough to warrant labeling.
Accessing via UUID, id or path is annoying because the names are really long. So I decided to create a set of udev rules enabling me to refer to drives based upon persistent nick names.
This way I can run fdisk /dev/fred instead
of fdisk
/dev/disk/by-id/scsi-HITACHI_HUS153030VLF400_J8V1WDXA_E4TFL4M1NGDE4TH.
Somewhat more convenient. At least to the humans among us.
Last week was crazy! First the SNIA Summer Symposium in San Jose where we had another face-2-face in the data integrity technical workgroup. Friday, I flew to Ottawa for the Linux Symposium.
- Linux Data Integrity Extensions - a reprint of my paper from the Linux Symposium proceedings.
- Slides from my presentation.
Also, as of yesterday my data integrity infrastructure has been merged and will be in 2.6.27-rc1.
Sat, 14 Jun 2008Perl is like poutine. It's bad. But so good. And bad. But good. Mmmhm.
Sun, 08 Jun 2008
Christmas Came Early This Year
Sun, 25 May 2008
Bye Jimmy. So long and thanks for all the music...
Fri, 11 Apr 2008
My blog appears to be turning into a food showcase here in April. I am really busy at work and consequently I have nothing else to report. Codito, ergo sum...
Sun, 06 Apr 2008
Sunday morning rolls.
- Books:
- Pacific Modern by Raul A. Barreneche
- The Green House by Alanna Stang & Christopher Hawthorne
One of the only positive side effects of being allergic to most industrially prepared food is all the awesome home baked goods I get to eat. After I make them, that is...
Needed a break from coding tonight and made rugbrød, crispy oat cookies and brunsviger cake.
Fri, 28 Mar 2008
No More Involuntary Anti-Aliasing
My desktop at the office is a Dell OptiPlex GX620. It originally came with analog output only, so I invested in a Dell DVI Adapter (aka an ADD2 or SDVO card).
For a long time I used the machine with my 20" LCD panel @ 1600x1200. When that conked out after years of faithful service I got a 24" display which does 1920x1200. Unfortunately, the SDVO card in my OptiPlex (a SiI 1364)) maxes out at 1600x1200 or 165MHz. So for a while I've had to settle for a craptastic analog signal that was blurry enough to send me down history lane to the long, sleepless nights in front of a Sun 3/60.
Being allergic to blurry fonts (and thus anti-aliasing) I set out googling for an HDMI SDVO card. Today I stumbled upon the following thread. I added the following modeline and my desktop is once again crystal clear. Yay!
(mkp@mojo) ~> cvt -r 1920 1200 # 1920x1200 59.95 Hz (CVT 2.30MA-R) hsync: 74.04 kHz; pclk: 154.00 MHz Modeline "1920x1200R" 154.00 1920 1968 2000 2080 1200 1203 1209 1235 +hsync -vsync
PS. As long as we're using stone age ~100dpi displays I'm sticking with -b&h-lucidatypewriter-medium-*-*-*-12-*-*-*-*-*-iso8859-1.
Mon, 24 Mar 2008I keep reading all these horror stories about Dell's support but have never had the need to interact with them despite owning many pieces of Dell equipment.
Last Wednesday my main test machine stopped responding. I went to the lab and found the friendly blinkenlights emitting Your-System-Is-Toast messages.
I tried to pull the fibre channel cards and disconnect the internal drives. No go. Checked all connectors and fans. Nothing. I also tried a different power circuit just in case my power relay was acting up. But the machine remained dead.
So I called Dell Support. Less than 5 minutes on hold and then a guy named Chris picked up. I gave him the service tag of the machine and provided a concise rundown of all the things I had tried. In return he didn't bother asking me if my monitor was turned on and whether the power cord was plugged in.
We spent less than a minute walking through the support wizard and he apologized for having to spend time on this, given that it was pretty obvious what the problem was.
Chris promised to send a tech to fix the machine the next day. Now, my lab is a few blocks away in an unmanned location. And it's cold a noisy to sit there the all day waiting for somebody to show up. So I asked if they could send the the new power supply and have me install it myself. No problem! Shipping to my home address saving me a roundtrip to Oracle's Montreal office wasn't a problem either. Within a few minutes he had three-wayed a courier rep into the call to confirm shipping details.
As promised, the courier showed up the next morning with a new power supply and (just in case) and replacement motherboard. Inside the motherboard kit was a everything I'd need to swap the boards, including thermal paste and syringe. However, just replacing the power supply did the trick. I let the machine run the supplied diags CD and since it completed without a burp I left the original system board in place.
Included in the shipment were two prepaid UPS waybills for returning the defect PSU and the motherboard. A note told me to call an 800-number and arrange pickup. I did so today. The courier stopped by within a few hours, and the parts are now on their way back to Dell. I left a note in the motherboard box that I hadn't touched it to save them the trouble of figuring that out.
That's the way to do it. Thanks, Dell!
Sun, 09 Mar 2008
I think it'll be easier to get potatoes from the supermarket...
Sun, 02 Mar 2008
We made perogies tonight. Not quite as good as Baba's but not bad for a first take.
Fri, 29 Feb 2008
Linux Storage & Filesystem Workshop
Back from the Linux Storage & Filesystem Workshop in San Jose and a quick visit to Oracle HQ.
At LSF08 I did a presentation on the data integrity project.
I have also posted a formatted version of my article from Enterprise Open Source Magazine: Proactively Preventing Data Corruption.
Another great book: Right Hand, Left Hand by Chris McManus.
Fri, 08 Feb 2008I have a 6-bay fibre channel disk array that--unlike the rest of my SAN setup--doesn't take 15 minutes to fire up. So it's been extremely useful for quick and easy I/O testing.
This particular model has been discontinued for a while and it only supports 1 and 2 Gbps link speeds. I recently got a batch of brand new drives that support 4 Gbps operation. And unfortunately the array backplane turned out to be incapable of negotiating link speeds with them.
First I tried to download a manual. I filled out the web form and waited for the "guaranteed next business day response"... Nothing.
Then I contacted the vendor's tech support who has been fantastic at getting back to me in the past. This time: No response. Neither from the web form, nor via email. I know this is a discontinued product, but still...
I worked around the problem by powering up the array with an old 1 Gbps drive plugged in. Once things settled down I hotplugged the 4 Gbps drives which happily went to 1 Gbps link speed. I never got the same stunt working with a 2 Gbps "bootstrap" drive. So I had to live with 1 Gbps.
But the main downside to this workaround was that I had to be in the lab every time I needed the array. Thus defeating the quick and easy criteria that was my main reason for using it in the first place.
Yesterday, I stumbled upon a manual for the array. I used the amazing powers of jumper W5 to force backplane to speed 2 Gbps. And presto! It worked.
I should point out that when I initially got the 4 Gbps disks I had tried several jumper combinations in an attempt to force link speed. But between backplane and I/O daughterboard there are many combinations.
All I really needed from the vendor's tech support was two letters: "W5". Was that really too much to ask? And why on earth don't they just put their manuals up for download?
Mon, 04 Feb 2008Some of the books I've enjoyed recently:
- Why Things Break by Mark E. Eberhart
- Arthur the Dragon King by Howard Reid
- Isaac Newton by James Gleick
- Why Beauty Is Truth by Ian Stewart
- The New Ecological Home by Daniel D. Chiras
This week I went to the SNIA Winter Technical Symposum in San Jose to talk about the work I've been doing on block level data integrity in Linux.
Last night I took CalTrain up to San Francisco to see an organ show at Pearl's. Hooked up with Pete Fallico and an incredibly squeaky B-3. I also met Matt, and after the show Pete gave me a ride back to my hotel in San Jose. It was like the good ol' days, except...
Back in the day I used to go between Menlo Park/Palo Alto and the Linuxcare HQ at 650 Townsend all the time. When I got off the train at 4th and King last night I thought I'd landed on a different planet. The station was the same. But there were - uhm - other buildings around it. The whole area south of the tracks has been developed.
<LouisBalfour> Crazy! </LouisBalfour>
It was also interesting going up to 3rd and Market. My old home-away-from-home, The Argent Hotel, is now a Westin.
The Times They Are A-Changin'
Sat, 05 Jan 2008| TP | 360C | 560 | 600 | 600X | X22 | X61s |
|---|---|---|---|---|---|---|
| Clock (MHz) | 33 | 100 | 233 | 650 | 800 | 1800 |
| RAM (MB) | 4 | 40 | 64 | 256 | 384 | 2048 |
| Disk (MB) | 170 | 810 | 4096 | 6144 | 20000 | 100000 |
| Display | 680x480 | 800x600 | 1024x768 | 1024x768 | 1024x768 | 1024x768 |
| Year | 1994 | 1996 | 1998 | 2000 | 2002 | 2008 |
← back