Today's mostly trivial and "well-known except if you haven't seen it before" system administration tip is prompted by one of the recent comments over at James' blog.

Be careful when using shell wildcard patterns as arguments to scripts. There are two ways to mess up and you'll usually forget to test both of them.

(Read more...)

Topics: software/linux, technology/sysadmin

[#]

Things that happened at the conference since I last wrote. I flew back from Melbourne last night, missing out on the Open Day today, so this is only notes from the final day and some overall impressions.

Also, the talks are gradually being encoded and put up on the conference website in both video and audio format. I gather there are plugins around to let Windows users play ogg-style video, although I don't have any experience with it. Not sure if it's a case of "not yet" or a blunder that only half of my "stand up and blather" session is available, but it's hardly the most interesting thing there and it's impressive how fast the videos can be made available.

(For those wondering, "most interesting" might be Bdale Garbee's rocket talk, or Jon Corbet's State of Linux talk, or Stormy Peters' keynote or possible one I haven't managed to watch yet.)

(Read more...)

Topics: software/linux, conferences/linux.conf.au

[#]

Melbourne airport's Qantas Club has free wifi, which is nice. However, all is not entirely sane in the network land:

...
Feb  1 18:24:36 lancre dhclient: bound to W.X.Y.Z -- renewal in 59 seconds.
...
Feb  1 18:25:35 lancre dhclient: bound to W.X.Y.Z -- renewal in 55 seconds.
...

and so on (public IP address changed to protect the innocent). Rinse, wash, repeat. I'm all for not tying up the network with unexpired leases that aren't being used, but 60 seconds seems a little eager. I'm just saying.

Topics: software/linux, conferences/linux.conf.au, technology/sysadmin

[#]

[I'm quite behind on a lot this month, due to a bit of unforseen downtime, but I want to get this down whilst it's fresh.]

Currently at linux.conf.au, feeling a bit rushed, but enjoying it now I'm here.

Have my tutorial this morning and it seemed to go reasonably well. I realise as I was writing the slides last week that, as usual, I'd vastly overestimated the amount of stuff I'd be able to cover in 90 minutes. I miss the three hour tutorial slots we used to have at this conference. Still, enough people showed up that they were sitting in the aisles: always a good stroke for the ego.

My talk was nominally about website server performance and I tried to structure things to give people a sense of the problems they have to identify and face. The entire field is pretty large, so I stayed away from too many specifics and concentrate on the attitude and approach that is required to understand and solve problems. Although the slides won't make a lot of sense if you weren't at the talk, they're available online for those who did attend and want to remember which tools I mentioned.

The talk was video taped and they gave me a microphone this year (vast improvement on last year), so it should be available online eventually.

Somebody asked, quite reasonably, if I blogged about the stuff I was talking about and, indeed, I should do more of that stuff. There are a lot of examples of good problem solving, both pre-emptive and after the post facto changes that sites have made, both large and small. I'll try to write some articles about that sort of thing over the coming months. Sharing experiences in the sysadmin/operations arena seems to be a good way to get the knowledge spread around.

The conference generally has been fun, although I only flew down yesterday (missing the mini-conferences on Monday and Tuesday). It was nice to hear Stormy Peters speak at the keynote this morning. Not sure I entirely agree with her conclusions, but her groundwork was sound and it provided food for thought. Val Henderson's talk on her clusterfs project was worthwhile, too. She's a good presenter and filesystem theory and practice is full of interesting algorithms and data structures.

Oh, I was lucky enough to get an OLPC in the giveaway yesterday, so I now have a nice little XO to play with.

More later.

Topics: software/linux, conferences/linux.conf.au

[#]

Rusty Russell's Hard To Misuse Interface Levels

Lower numbers are better. Really. Source is here. See below for details.

1. It's impossible to get wrong.
2. Compiler/linker won't let you get it wrong.
3. Compiler will warn if you get it wrong.
4. The simplest use is the correct one.
5. The name tells you how to use it.
6. Do it right or it will always break at runtime.
7. Follow common convention and you'll get it right.
8. Read the documentation and you'll get it right.
9. Read the implementation and you'll get it right.
10. Read the correct mailing list thread and you'll get it right.
11. Read the documentation and you'll get it wrong.
12. Follow common convention and you'll get it wrong.
13. Do it right and it will break at runtime.
14. The name tells you how not to use it.
15. The obvious use is wrong.
16. Compiler will warn if you get it right.
17. Compiler won't let you get it right.
18. It's impossible to get right.

(Read more...)

Topics: software/design, software/linux

[#]

Following my post last week about Flickr's interpretation of my photos' EXIF metadata, I spent a little more time this afternoon looking deeper into the problem.

End result: I understand the problem a bit more and have a workable solution...

(Read more...)

Topics: software/linux, photography

[#]

I had neglected to read Greg Kroah-Hartman's OLS keynote until just now. I've heard Greg speak quite a few times, so I can imagine him saying all this as I read it. Excellent stuff.

Talks about Linux kernel development have two benefits. They're usually just plain interesting, which is always nice. However, they also provide lessons for non-kernel projects. The Linux kernel development has been going on in public for a lot of years (almost 15, if some recollections from Linus are accurate, which they no doubt are). It's had peaks and troughs and the linux kernel mailing list (lkml) is a tough place to work. But it works! There is no more evidence required that they are doing something right. I suspect quite a reasonable theory of software engineering course could be built around the history of Linux kernel development. It would be interesting, too.

Topics: software/linux

[#]