There's some tuning knobs in the Application Settings section of the
admin console. You could try increasing the Min Pending Latency and
decreasing the Max Idle Instances.
I've done that. Doesn't seem to help. I have:
Max Idle Instances set at "Automatic-1" which is the lowest setting.
Min Pendency Latency set at "15-Automatic" which the the highest setting.
The scheduler still occasionally spins-up a second instance which then proceeds to go idle then eventually goes away. The problem is that it's needlessly using resources that I end up paying for. I enabled billing on Saturday and prior to tweaking these settings ended paying for additional instance hours. It's unclear with it tweaked this way whether I'll need to pay for additional hours. But, based on a causal monitoring it looks like it's going to be nearing the 28 hour limit most of the time. There really should be away to say I want a single instance and no more. It's a minor inconvenience at the moment because I'm unlikely to hit the monthly minimums in the near future so there is room at the moment for paying for additional but not needed instance hours. However, that may not always be the case.
There's also a feature request on the issue tracker for this:And, a relate issue that would allow a single instance the ability to handle more then 10 concurrent requests
The instances don't go away when Max Idle Instances is set at "Automatic-1", but you do stop getting charged for them, which should hopefully address your concern. You can verify this using the Instances chart (look at the "billed" line)