cloudRunStayAwake 1.0.1

Minimising Cloud Run cold start with targetted self call

Cloud Run Stay Awake

Python implementation of the cold start mitigation proposed by @guillaumeblaquiere (already implemented in Go & NodeJS )

Purpose

A Cloud Run GCP service stop after an idle period, this time cannot be forseen. As a result, the next request after the stop, will have a cold start, a delay of one or several seconds, Cloud Run takes to start a new container. One of the cheap solutions proposed by Guillaume (https://medium.com/google-cloud/3-solutions-to-mitigate-the-cold-starts-on-cloud-run-8c60f0ae7894) is to schedule an http call every minute to keep alive the service. However this solution is not accurate as the stop can occurs in less than a minute after the last request. And Cloud Scheduling can't go under a minute.

Another solutions is presented to trigger the stop signal (a SIGTERM signal) to self call the service to try to keep it alive. This is what this code is about.

Implementation

In your code (Flask, Django, other...), two items must be included:

• signal.signal(signal.SIGTERM, stayAwake) in addition to the import signal (builtin library)
• Copy the stayAwake function from the main.py file, that will do the keep alive logic

Tests

In case you want to perform some tests before deploying to your code, there is a Dockerfile that can be used to test it on Cloud Run

License

This library is licensed under Apache 2.0. Full license text is available in LICENSE.