Daily usage has shot up, eating through thousands of instance hours

[Paul Canning]

We have noticed that in the past month, instance hours (front end and Cloud SQL) have shot up, in some cases tripling what it used to be.

We have 3 main app engine modules running

1. WordPress site
2. CodeIgniter based PHP web app
3. CodeIgntier based PHP custom API

We don't have a ton of traffic, but we do poll the API a lot (we have other web apps that poll the API to see if it can connect and run)

It is 10:40am here in the UK and already we have used 300+ instance hours for all of our app engine stuff.

What is going on?

We are not going over any of our quotas, it's just instance hours. So how do we keep that cost down?

[Evan Jones]

There are a ton of things that are different between your situation and what I'm familiar with, and I'd want to see a ton more details to actually have a ssuggestion. However, as one random thing that we've seen make a difference: Adjusting the "max_idle_instances" seems to impact billing pretty strongly for our application. At this point, we run nearly all our modules with a low value for max_idle_instances (e.g. 1), since it seems to decrease cost a fair amount, and haven't observed a latency impact (we also haven't monitored it TOO closely and we aren't super latency sensitive). The way we see this as a problem: On the App Engine dashboard, if you select "instances", watch the "billed instances" VS the "active instances". In our case: we see "billed" to be usually below "active." When this was a problem, it was higher.

Good luck, and if you figure out what was happening, I'd love to know (a) what and more importantly: (b) how you tracked it down.

Evan