This is a template project for creating new PushPy Raft-based Python application server deployments. Each deployment is different will likely have different boot and configuration setups. Base Docker images are provided to simplify customization.
The key components are the
- boot - the boot python module that initializes pushpy
- conf - the config setups for different pushpy nodes
The docker image scripts are setup to use these two directories when populating the image.
The deploy directory has two examples of deployment
- bare
- simple-cluster
The bare example shows how to deploy pushpy without containers which can be useful on small machines (e.g. Graviton). The simple-cluster shows how to deploy using containers and emphasizes the bootstrapping style where there is a primary node and secondary nodes are launched from it.
The docker directory shows how to create custom pushpy images that are derived from the base pushpy image. There are two example images
- python3.8
- tensorflow
These are obviously just different deployments, with the tensorflow image being optimized for machine-learning.
Quickstart examples to bring up a pushpy cluster using two different strategies.
# terminal 1
$ source setenv.sh
$ cd deploy/simple-cluster
$ ./primary.sh python3.8 up
# terminal 2
$ source setenv.sh
$ cd deploy/simple-cluster
$ ./secondary.sh python3.8 up
# terminal 3
$ source setenv.sh
$ cd deploy/simple-cluster
$ ./secondary-2.sh python3.8 up$ source setenv.sh
$ python3 -m venv venv
# terminal 1
$ source venv/bin/activate
$ cd deploy/bare
$ ./run_primary.sh
# terminal 2+
$ source venv/bin/activate
$ cd deploy/bare
$ ./run_secondary.sh