Wednesday, June 13, 2007

08:23:17 AM: Mr. Wizard dies at 89:

Don Herbert, who explained the wonderful world of science to millions of young baby boomers on television in the 1950s and '60s as "Mr. Wizard" and did the same for another generation of youngsters on the Nickelodeon cable TV channel in the 1980s, died Tuesday. He was 89.

I loved watching Mr. Wizard and was part of the second generation that he explained science to. Definitely influenced my life-long love of science and decision to get a degree in physics.

08:26:20 AM: On the work-excitement vs. training-excitement spectrum I've definitely been leaning way towards work in the last week or so. Lots of good stuff happening and getting done. But I also have to keep up with the training sched. Today's long run day. Last week's long run day was dismal. Today I again feel excited only by getting off the run and back to the computer to implement some of the many check list items that are waiting for me. Not a good way to go into a long workout. So I'm procrastinating right now. Should be on the trail running already. Instead I'm working. If I wait too long I'll catch the afternoon heat. Need to get out there asap.

08:39:59 AM: Jason Calacanis, the actual CEO of Mahalo, posted a link to my little post about what I'd do as CEO of Mahalo. Thanks man! There's an extra space in the link right now and it's resolving to a 404 but I shot him an email and hopefully he'll be able to fix it soon. Good news is that they're listening and are launching Mahalo Greenhouse. Fifteen days since launch and new user-requested features are flowing. Certainly something to shoot for as I launch dneero.com.

09:38:47 AM: Still haven't left for the run.

10:44:58 AM: New dNeero Closed Beta code release. Some very cool new social features. Oh so close to launch. Can smell it. Just a little network testing that can now be done thanks to Sr's data center exploits yesterday. And then we're launching.

10:47:13 AM: So dNeero's something like a drug these days. I just can't get away from it. This is exciting. But I need to run long today. Should I commit to a distance, only to embarrass myself later on? Why the hell not? I'm thinking 23 miles. Most days wouldn't be a big deal but with the leg sticks on the cusp of overtraining I never have any idea what they're gonna feel like as I start a run. Really hoping to get out there soon because I'm already tempting the afternoon heat to roast me.

11:27:58 AM: At the trailhead. Finally going running.

03:36:15 PM: Done. Good run. 23 miles in 3:30. Saw Stewbeef out there riding.

05:38:57 PM: dNeero is now load balanced between two front-end servers each connected to a back-end database server. Seems to be working fine thus far but before launching I want to do some testing. This is a big step because it grants us a valid scaling strategy... pop in more front-end servers until we blow up the database server. In theory the front-ends are clustering/caching data to reduce load on the database server. But I haven't verified that the cluster properly formed yet. Like I said, more testing to do.

05:42:06 PM: And of course I spoke too soon. On redirects the Apache load balancer, which sits in front of the front-end servers, switches servers. The cluster doesn't have time to sync sessions meaning that the user who's logging-in gets bounced. Now, in theory Apache does something called sticky sessions which prevents this. But that seems to be failing right now.

10:39:29 PM: Quick trip to Walmart earlier in the evening for a rake, fan, iron supplement, pillow and some other even more random crap.

Then picked up a Chick-Fil-A milkshake for Terp and for me. Snugged with Aves as she went to sleep.

And since then have been working on the server clustering issue. Seems to be working again thanks to this little gem:

I had a problem like this, and determined that the jvmRoute value in my
server.xml did not match the names of the tomcat workers in the
workers.properties file. If your workers.properties file defines
workers, t1 and t2, and "worker.loadbalancer.balanced_workers=t1,t2",
then your server.xml file must use t1 and t2 as the value for jvmRoute=
in the tomcat configuration.

Which I didn't find in the documentation anywhere. Or I missed it. The jvmRoute attribute goes on the server.xml Engine node. So Apache httpd mod_jk sticky sessions appear to be working with Tomcat 5.5.

I know, I know. They'll break three seconds after I post this.

10:46:18 PM: Not broken, yet. But the Hibernate cluster isn't sharing objects across the wire, yet. This isn't optimal, and could be dangerous. I need to get them synchronizing. They did so on local test servers. But working when promoted to production is, of course, out of the question and takes all the fun out of it.

10:53:42 PM: Brandon Wren's up late beating on the dNeero Closed Beta. And he found a bug. Thanks Brandon!

11:40:52 PM: Heading to sleep. Today was action-packed. In a keeky triathlete sort of way.