The instructions here are for a multi-process version of the Python 2.7 runtime currently in alpha. Please do not use this version for production workloads. To use the stable version of the runtime, follow the documentation at https://cloud.google.com/appengine/docs/managed-vms/ instead.

• Using the multi-process runtime
• Configuration
• Building your own version
• Caveats

These instructions assume you have a working Python application that has already been deployed successfully to Managed VMs using the default runtime version.

As this version of the Python runtime is currently in alpha, it is not built as a stable Docker image like other runtimes. Instead it can be put into place via modifications to an application's Dockerfile.

If your application has an automatically generated Dockerfile (this will be created during the deployment process the first time your app is deployed), copy multicore_runtime/dev/Dockerfile from this repository to your application directory.

If your application has a custom Dockerfile, then compare the two files and merge them. In typical cases this can be done by removing the first line of your application's Dockerfile (starting with FROM) and the last line of multicore_runtime/dev/Dockerfile (ADD . /app) and concatenating the files, with the repository's version first.

This Dockerfile will download the latest release of the runtime directly from Github. If you would like to instead use a version of the runtime you have modified or that has not yet been released, see the "Building your own version" section.

Post-release, this version of the Python runtime will be made available as a prebuilt Docker image with no manual Dockerfile modification required.

By default the multi-process version of the runtime is launched via the Gunicorn webserver and is configured to use a fixed number of processes and gevent-based concurrency.

This can be changed by editing the ENTRYPOINT line in the Dockerfile. For instance, Gunicorn can be changed to Waitress or another Python webserver, or the number of processes can be increased, or gevent-based concurrency can be turned off.

A future version of this runtime may automatically adapt the default settings to accommodate a given instance size. This version instead uses a fixed number of processes. If you are using an instance type with more than 4 CPU cores, change the number of processes in the Dockerfile to between 1 and 2 times the number of CPU cores available.

If you would like to make modifications to the runtime (either for personal use or to debug or resolve an outstanding issue), you can build and deploy a custom version with the following steps:

• After cloning the repository locally and making your changes, build a source distribution by running python setup.py sdist from the root directory of the repository.
• Copy the resulting tar.gz file to your application's folder, in the same directory as the Dockerfile.
• Edit the Dockerfile in your application and look for the line that says ADD (...) /home/vmagent/python-runtime.tar.gz
• Replace that line with a COPY command, with the filename of your generated tar.gz file instead of the URL as the source, and the same destination. An example is included in the Dockerfile comments.
• Deploy your application. A warning during deployment where your tar.gz file is rejected for addition because it is too large can be ignored.

As this is an alpha product, some functionality has not yet been implemented.

• Logging is via stderr only. The default logging view in Google Developers Console is "Request". Click on the dropdown showing "Request" and select "stderr" to view the logs. stderr-only logging is temporary and will change in a future version.
• Log entries are not automatically associated with a specific request.
• Handlers that are flagged as login: required or login: admin are not supported. Attemping to access these handlers will result in a 404 as the handlers will not be registered.

There may be other features that work on the current Python runtime and are not implemented or not functional in this version. Please open an issue on Github for any you encounter.