Doesn't the App Engine free tier include southamerica-east1?

[Diego C.]

Hello,

I had an App Engine service running on us-east1 for months. It was always within the free tier, so I was never charged anything.

A couple of days ago, I shut down that project, created a new project, and started an App Engine service on southamerica-east1. Just a testing app, very little traffic, never went over the free tier limits.

Today I got a billing alert saying that I had reached my budget limit of $2 that I set a while ago, precisely for cases of unexpected charges. The billing details show that I was being charged for App Engine frontend instance hours on southamerica-east1 (São Paulo). I shut down the project to prevent any further charges.

(Note that I never ran both services concurrently.)

Why was I charged for the same thing in southamerica-east1 but not in us-east1? I haven't found any mention of the App Engine free tier being limited to certain regions, as it is for Compute Engine, for example.

Thanks!

[Mary (Google Cloud Support)]

Hello Diego,

The App Engine free tier usage is covered in[1], as you can see it covers 28 frontend hours and 9 backend instance hours per day as well as other services, however this only applies on App Engine Standard.  

The instance hours begin when an instance starts and ends depending on the type of scaling you specify for the instance. Once the free limit has been reached, depending on your region you will be charged the rate for your instance type as seen here[2].

[1] https://cloud.google.com/free/docs/gcp-free-tier#always-free-usage-limits

[2] https://cloud.google.com/appengine/pricing#standard_instance_pricing

[Diego C.]

Hello, Mary.

Thank you for your reply.

I did some further research. I was in fact charged for more than 28 instance hours per day for 3 days, starting on April 4 (when the service was started) and ending April 6. I shut down the project around noon on April 7, so it didn't go over the quota that day. For example, on April 5, it was 38.94 hours:

However, I restored the project so I could take a look at the dashboard, and this is the chart of instances used:

You see that it spends most of the time running 1 instance (I set min_instances to 1 in app.yaml) with a few spikes which were apparently caused by old versions starting up an instance for a short time every once in a while. However, that doesn't account for over 10 extra hours on April 5 alone, and even more hours on April 4, which only had one spike.

I had no other projects using App Engine.

What am I missing?

Thanks again!

Diego

[Diego C.]

OK, after more research, I think I know what happened.

When I first started this service, I deployed two versions with errors that crashed the app as soon as it started. As I said, I was just testing things. I deployed a properly working app within a few of minutes. However, judging by the logs, it seems like App Engine continuously tried (and failed due to the errors) to launch instances of the old versions, at a rate of around once per second per version, for days. These old versions had no traffic routed to them. I suppose that these fractions-of-a-second-per-second that the instances were running, eventually added up to several hours per day. The instances were up for such short times that they don't appear in the chart, except for the occasional spike.

I guess it's my fault for deploying apps with errors, but I think that this definitely could be handled better by App Engine. One doesn't expect non-functional, old versions to be kept running.

Diego