Based on remote-docker-aws 🙏🏾
Use docker to develop services, but without the overhead of running docker on your machine! This is a development tool that you should use if your machine is low performance, or if you are running many docker services.
Frees up your local machine for useful tasks such as running your code editor, browser, and email, leaving running Docker to a dedicated server instance. The result is that your local machine functions faster, uses up less disk space, and consumes less power. MacOS users will also see noticeable speed improvements since Docker on Linux (which is what the remote hosts runs) is much more performant.
The downsides:
- SSH tunnel communication is slower than local communication. However using an AWS region with low ping makes the latency unnoticeable. Find the region fastest for you using this site
- Some more setup required to get everything configured properly and running (tunneling ports, syncing file changes)
- Running the ec2 instance incurs an additional cost over running locally, although a t3.medium instance in Canada only costs just under 5 cents/hour
How it works: two processes are run, a sync and a tunnel process.
- The sync process keeps local and remote files in sync so that the docker process run remotely can use docker volumes transparently
- The tunnel process forwards ports needed so your local system can communicate with docker, plus additional ports as required, such as port 443 for browser communication
-
First login to your AWS account and create access keys to access AWS through the CLI
You will need the following IAM policies:
- AmazonEC2FullAccess
- AWSCloudFormationFullAccess
And now in your terminal:
# Replace dav with your name # You will need to setup an AWS account if you don't have one # and create access key credentials aws configure --profile dav export AWS_PROFILE=dav
-
Install pre-requisites
Have Homebrew (Available on both macOS and Linux now!)
Have pipx
pip install perry-the-docker-agent # or.... if part of a project poetry add perry-the-docker-agent # Install unison sync utility brew install unison # Install file-watcher driver for unison # On MacOS: brew install autozimu/homebrew-formulas/unison-fsmonitor # Or, on Linux since the above formula doesn't work: brew install eugenmayer/dockersync/unox
-
Generate and upload a keypair to AWS
perry create-key-pair
-
Create the ec2 instance
perry create
-
Start the remote-docker ec2 instance
perry start
This will automatically switch the docker context for you. If you want to switch back to the default agent run
docker context use default
-
In one terminal start the tunnel so that the ports you need to connect to are exposed
perry tunnel
-
In another terminal sync file changes to the remote instance:
perry sync
-
Develop and code! All services should be accessible and usable as usual (eg:
docker ps
,docker-compose up
, etc.) as long as you are runningperry tunnel
and are forwarding the ports you need -
When you're done for the day don't forget to stop the instance to save money:
perry stop
Looks for a config file at the path ./perry_config.yml
by default,
which can be overriden by passing --config-path
.
An example perry_config.yml
file:
aws_region: af-south-1
instance_type: t3.large
volume_size: 100
sync_dir: ./
project_id: "ranger"
ignore_dirs:
- .venv
- .git
- node_modules
- __pycache__
remote_port_forwards:
local-webpack-app:
"8080": "8080"
local_port_forwards:
user-api:
"2020": "2020"
blog-api:
"3030": "3030"
Usage: perry [OPTIONS] COMMAND [ARGS]...
╭─ Options ───────────────────────────────────────────────────────────────────────────────╮
│ --config-path TEXT Path of the perry config │
│ [default: ./perry_config.yml] │
│ --install-completion Install completion for the current shell. │
│ --show-completion Show completion for the current shell, to copy it or │
│ customize the installation. │
│ --help Show this message and exit. │
╰─────────────────────────────────────────────────────────────────────────────────────────╯
╭─ Commands ──────────────────────────────────────────────────────────────────────────────╮
│ create Provision a new ec2 instance to use as the remote agent │
│ create-key-pair Create and upload a new keypair to AWS for SSH access │
│ delete Delete the provisioned ec2 instance │
│ ssh Connect to the remote agent via SSH │
│ start Start the remote agent instance │
│ stop Stop the remote agent instance │
│ sync Sync the given directories with the remote instance │
│ test Test new config │
│ tunnel Create a SSH tunnel to the remote instance to connect with the docker │
│ agent and containers │
╰─────────────────────────────────────────────────────────────────────────────────────────╯
The current configurable values can be seen in config.py
- The region to create the instance in
- Type of ec2 instance, defaults to:
t3.medium
- defaults to
None
- Used to uniquely identify the instance, this is useful if multiple remote-docker agents will be created in the same AWS account
- defaults to:
~/.ssh/id_rsa_perry_{{project_id}}
- defaults to:
{}
- Object containing label -> port mapping objects for opening the ports on the remote host.
A mapping of
"remote_port_forwards": {"my_app": {"80": "8080"}}
will open port 80 on your local machine and point it to port 8080 of the remote-docker instance (which ostensibly a container is listening on). The name doesn't do anything except help legibility.
-
defaults to:
{}
-
Similar to
local_port_forwards
except will open the port on the remote instance.This is useful to have frontend webpack apps accessible on the remote host
- defaults to:
[]
- list of directories to ignore
- directory to sync, will usually be the root fo the project
- defaults to:
30
(GB) - Size of the ec2 volume.
A t3.medium instance on ca-central-1 currently costs $0.046 /hour. See current prices
Nothing else used should incur any cost with reasonable usage
- See
perry --help
for more information on the commands available - The unison version running on the server and running locally have to match. If one of them updates to a newer version, you should update the other.