deployment

Pull Git repositories and run each one's deploy script

TLDR

Watch `/etc/deployments.csv` and deploy into `/var/deployment`

$ deployment

Use another list and another home directory

$ deployment -d [path/to/deployments.csv] -H [path/to/directory]

Clone anything that is missing, then exit

$ deployment -c

Print the commands without running them

$ deployment -n

Wait this many seconds after each repository

$ deployment --delay [60]

Copy every repository `dist` directory under the home `dist` directory

$ deployment -D

Write a health timestamp after a fully successful pass

$ deployment --health-file [/tmp/deployment-health]

SYNOPSIS

deployment is a long-running Python program that keeps a list of Git checkouts up to date and runs a `deploy` script at the root of each one. It is aimed at publishing sites and other build products. Python 3.9 or newer, git, and rsync are required. There is no distro package. The usual install is to copy `deployment.py` onto `PATH` as deployment and run it under systemd or in a container that includes the build tools those scripts need.Each CSV row is `path,repository_url`. A path that starts with `/` is absolute. Any other path is relative to the home directory. A checkout with no executable `deploy` script is only pulled. A `deploy` script is arbitrary code: chmod +x it, and do not point the list at a repository you do not trust.The published container image (`chrissx/deployment`) expects the list at `/etc/deployments.csv` and data under `/var/deployment`, and it turns on `--health-file /tmp/deployment-health`. The image is based on Ubuntu and includes Node. It is not a general-purpose build environment.

PARAMETERS

-d, --deployments file

CSV of deployments. The default is `/etc/deployments.csv`

-H, --home dir

Base directory for rows whose path does not start with `/`. The default is `/var/deployment`

-c, --clone-only

Clone repositories that are not on disk yet, then exit. Do not pull or build

-n, --dry-run

Print commands instead of running them. This does not predict rsync or what --global-dist would copy

-D, --global-dist

After a build, copy that repository's `dist` directory into the home directory's `dist` directory

--delay seconds

Pause after each pull and build. The default is 30. The first pass does not wait

--health-file path

After a pass that succeeds for every row, write an expiry timestamp to this file. Startup and deploy errors remove the file. --dry-run and --clone-only leave it alone. An empty list never counts as healthy. The expiry budget is the number of rows times --delay times 2

CONFIGURATION

/etc/deployments.csv

Default deployment list. One `path,repository_url` row per checkout.

/var/deployment

Default home for relative paths, and the place --global-dist collects `dist` trees.

deploy

Executable script at the root of a repository. Typical contents install dependencies and run the project's build. If the file is missing, deployment only runs `git pull`.

CAVEATS

`deploy` scripts run with the privileges of the deployment process. Do not use this tool on untrusted repositories. Build tools can leave zombie processes. A container should use an init process so those children are reaped.--dry-run cannot show the full effect of rsync or --global-dist. --health-file stays untouched during a dry run or a clone-only pass, and an empty CSV never becomes healthy.The name is a generic English word. This page is the chrissxMedia deploy loop, not a Kubernetes or CI product that happens to be called deployment.

HISTORY

deployment is maintained by chrissxMedia as a small Git-driven publisher for their own sites and for projects they host. The Docker image that pairs it with Node 24 is newer than the systemd unit and is still lightly tested upstream.

RESOURCES

Source code