Yesterday, I decided to upgrade one of my development boxes to the new Fedora Cora 2 test. Fedora is what Red Hat is evolving their personal distribution into, and the new Fedora Core 2 test is the first distribution to include a version of the new 2.6 kernel. I use two boxes for development. I have one machine, called floorpie, that I basically use as a terminal. I do my development on a second, much more up-to-date dual-processor box. Both machines had been running RedHat 9. The 2.4 kernel has always bugged me, because it never scheduled things well, at least in terms of a desktop OS. I generally run XMMS and listen to music while I type away. But under 2.4 whenever I would move cursors between windows, XMMS would stop for half a second or so, along with the music. Very distracting. And it’s not like the box is underpowered or anything. With 2.6’s vastly improved scheduler, I figured that was my ticket to uninterrupted music. So, using the most excellent Yum upgrade system, yesterday I upgraded floorpie first to Fedora Cora 1, and then to the test version of Fedora Core 2. Many hours of debugging/fiddling later, I have a mostly working system. I ran into a couple of big problems. First, yum didn’t pull down a binary version of the new kernel. I didn’t notice this until I rebooted, and found myself still in a 2.4 kernel left over from Fedora Core 1. I also found myself without networking. Yum did install the source RPM for the new kernel, so I went and compiled a 2.6 kernel using it. That went fine, and I was able to reboot into 2.6. Next, I had to tackle the network problem. For some reason, the modprobe.conf file was changed to reference a different network module. I finally found that and changed it back to reference the correct module (the tulip driver in this case), and I was able to get networking working again. Next up was sound. Floorpie has a SBLive chipset. When I tried to play something, all I’d get was a low level of static. After much fiddling around, I found that the driver was correct, but the mixer levels were all screwed. Somehow, my copy of Gnome currently doesn’t include a graphical mixer program, so after some fiddling, I found alsamixer, which is a curses-based mixer. After much fiddling, I again have sound. There are still a couple of problems for which I have not found solutions. The new Gnome has focus problems. Previously, to pop a window to the front, I could just click anywhere on the window. Now, I need to hold down the Alt key while doing that. Best I can tell, this is a bug in Gnome. Hopefully it’ll be fixed soon. The other problem is with X resources. I run emacs from my main development box and view it on floorpie. I have a resources file to change the colors. For some reason, when running emacs remotely, those resources are no longer being used. I can’t figure out why. They still work if I run emacs locally. So those are the problems, but did the new kernel fix the skipping music problem? Yes, absolutely. The new scheduler appears to work very well. The desktop is snappy in all regards. Based on some benchmarks I’ve seen, I can’t wait until we can start using the new kernel in production for Bloglines. My guess is that that’ll be in about 6 months. For now, I just need to find solutions to the remaining problems with Fedora Core 2 test.

I’m confused by something I’m seeing with Technorati. If I search on www.bloglines.com, it says that there are ‘1106 inbound blogs’. So then I go to the Top 100 listing, Bloglines doesn’t show up. If I’m reading things corrected, we should appear at about number 64, right after Gawker. But we’re nowhere. Strange. Maybe I’m misunderstanding something?

Back when we were going through the ONElist/eGroups merger, we had some fairly significant downtimes, as we first moved the ONElist service to our own servers from Digital Nation, and then as we merged the two services. We needed a down page, something to display on the web site during these down times. I wanted it to be something humorous, and I worked with Jack Wu to create what became the ’eGroups Plumber’. Tomorrow, we are moving the Bloglines site from one data center to another, and upgrading some of the hardware along the way. The site itself should only be down for about an hour, if we’ve done our homework correctly, but we still need a down page. So, I proudly announce the return of the plumber.

I’ll always have a soft-spot for the Merc. David Plotnikoff wrote a great article about ONElist back in the day. Mike Langberg writes today in Software keeps track of Weblog, news-site updates that:

If you want to stick your toe in the RSS water, I’d recommend starting with Bloglines, because it’s free and there’s no software to install on your computer.

User testimonials are always the best PR.

Bloglines has been sluggish at times the past couple of days. Our user numbers and web traffic keep increasing at a good clip, and we reached a point where the existing hardware couldn’t handle it. So we added more hardware yesterday and moved one of the databases around. That helped with the load, but it uncovered another bottleneck, which we were dealing with today. And then this evening, our main web server died. I haven’t had a chance to go to the co-lo, so I don’t know what’s wrong with it other than it’s down and isn’t coming back up. We have backup machines, and were able to switch over to one of them (we’ll be running on multiple web machines when we switch co-los in another week, but not at the moment), but the site was off-line for about half an hour. In addition, the favicons (the little icons next to the subscriptions), are off-line for the moment. We’ve spent a lot of time designing the Bloglines system to be scalable, and it is. But sometimes it’s difficult to simulate actual loads and therefore sometimes bottlenecks slip through, only to be uncovered on the running system. Scalability is both a science and an art. Scaling a system is a good problem to have, it means that people are using the service. And they are. But it can take some tweaking to get right. You spend time before launch to design a good system, then you run like hell after launch to keep up with the growth. So, thanks to everyone who uses Bloglines, and I apologize for the system issues. Hopefully they aren’t too noticable and they should be cleared up soon.