Little Plugins Can Have Odd Results

Some time ago, I installed something called Viral Headlines on the Allpar news site. The idea behind that is that you put in different headlines, and sees how people respond to them.

The only problem is, the coding of the plugin really, really caused a lot of work for the database. And by “a lot of work,” I mean changing its use of the processor time from 8% to 660%. (When you have eight processors, 800% is the maximum.) There were dozens of threads waiting for the table to be unlocked (reminder to self, when nobody is looking, take the server down, switch that table to something that’s not vulnerable to table locks.)

allpar-news-400x60

In human terms, it made looking at Allpar News much slower. People couldn’t get in.

The plugin is a great idea, but poorly implemented. I wonder about the coding. How did it vary the headlines when our headlines are cached everywhere? How did it get data that changed minute to minute? I can’t figure that out.

I need to be more careful what I install. Not everything is what it claims to be, even in the open source world.

I took Viral Headlines out, restarted the database to clear out all those old threads, and waited, and now we’re back at 8%. [sigh] And here I was wondering why the load was so high yesterday; I should have investigated.

Life lesson #1: investigate when you see something that’s not right in any way, no matter how minor. I should know that. In the survey business, I really learned it the hard way, but sometimes I forget to apply it elsewhere.

Life lesson #2: if I do install something new, watch it very closely for a while.