Closed mjkelly closed 10 years ago
It hasn't blown up spectacularly yet. We're on track to only use about 24 instance-hours today, which will be $free.
I've merged branch 'threadsafe' into 'master' and I'll close this. I can only hop we'll become so popular that we'll have this problem again soon.
Over the last month, there have been a lot of days we've been billed for 35-45 instance-hours per day. (Right now, for instance, I see 4 instances running simultaneously.) Looking at graph of instances over time, it looks like we're running a few pretty much continuously.
Tried setting the pending latency to 500ms (under Admin > Application Settings > Performance), which was quite a bit greater than the latency of most requests I'm seeing. It's back at 4 instances, after I killed 2 of them.
Next thing to try I think is setting threadsafe = true. I started a 'threadsafe' branch. I've done the CGI to WSGI, and looked through the code to try to spot obvious thread-safety problems (didn't spot anything).
I've set the 'threadsafe' instance as default. I'll see how it goes tomorrow. 'beta' is the old instance name, if we need to revert.