feat: add reusable GitHub Action to generate ImageCatalogs (#323)

Introduces a composite action that wraps `catalogs_generator.py` to
generate CloudNativePG ImageCatalog YAMLs from a container registry.
Supports multiple image types, distributions, and custom family prefixes.
Generates a `kustomization.yaml` for easy deployment of all catalogs.

Related to  cloudnative-pg/postgis-containers#100

Closes #324

Signed-off-by: Niccolò Fei <niccolo.fei@enterprisedb.com>
Signed-off-by: Gabriele Bartolini <gabriele.bartolini@enterprisedb.com>
Signed-off-by: Marco Nenciarini <marco.nenciarini@enterprisedb.com>
Co-authored-by: Gabriele Bartolini <gabriele.bartolini@enterprisedb.com>
Co-authored-by: Marco Nenciarini <marco.nenciarini@enterprisedb.com>
This commit is contained in:
Niccolò Fei
2025-09-24 18:14:03 +02:00
committed by GitHub
parent 607f425c34
commit 15d90eebd4
4 changed files with 296 additions and 29 deletions

View File

@@ -0,0 +1,117 @@
# Image Catalogs Generator Action
This composite GitHub Action generates [CloudNativePG ImageCatalogs](https://cloudnative-pg.io/documentation/current/image_catalog/)
from a container registry.
It wraps the [`catalogs_generator.py`](./catalogs_generator.py) script and makes it easy to
run inside CI pipelines.
---
## How it works
1. The script retrieves all image tags from a container registry.
2. A regular expression is applied to select the tags to include in the ImageCatalog.
3. Matching tags are sorted using [semantic versioning](https://semver.org/).
4. For each PostgreSQL major version, the latest matching tag is chosen.
5. The action generates:
- One `ClusterImageCatalog` YAML file per requested distribution and image
type
- A `kustomization.yaml` to install/update all cluster catalogs at once
---
## Inputs
| Name | Required | Description | Example |
| --------------- | --------- | ------------------------------------------------------------------------------- | ----------------------------------- |
| `registry` | ✅ yes | The container registry to query. | `ghcr.io/cloudnative-pg/postgresql` |
| `image-types` | ✅ yes | Comma-separated list of image types. | `minimal,standard` |
| `distributions` | ✅ yes | Comma-separated list of supported OS distributions. | `bookworm,trixie` |
| `regex` | ✅ yes | Regular expression used to match image tags. | *See [Regex](#regex)* |
| `output-dir` | ✅ yes | Directory where generated catalogs will be written. | `./` |
| `family` | ❌ no | Family name for generated catalogs (filename prefix). Defaults to `postgresql`. | `my-custom-family` |
---
## Regex
The `regex` input defines which tags are added to the `ClusterImageCatalog`.
- The **first capturing group** must be the PostgreSQL major version:
- `(\d+)` → e.g. `18`
- Subsequent capturing groups are optional and may include:
- an additional version: `(\d+(?:\.\d+)+)` → e.g. `1.2.3`
- a 12 digit timestamp: `(\d{12})` → e.g. `202509161052`
**Examples:**
```regex
# Matches '18-202509161052', '18.1-202509161052', etc.
'(\d+)(?:\.\d+|beta\d+|rc\d+|alpha\d+)-(\d{12})'
# Matches '18-3.0.6-202509161052', '18.1-3.0.6-202509161052', etc.
'(\d+)(?:\.\d+|beta\d+|rc\d+|alpha\d+)-(\d+(?:\.\d+){1,3})-(\d{12})'
```
> **Note:** Each `image-types` and `distributions` will be combined together
> to form a suffix, `-<img_type>-<distribution>`, which will internally be
> appended to the `regex` provided. Tags that do not contain explicit
> image type and distribution as a suffix are currently not supported.
---
### Family
The optional `family` input customizes:
1. **File prefix**: `<family>-minimal-trixie.yaml`
2. **`metadata.name`** in the ImageCatalog: `<family>-minimal-trixie`
3. **`images.cnpg.io/family` label** on the ImageCatalog object
If not specified it defaults to `postgresql`.
---
## Usage
Example workflow:
```
jobs:
generate-catalogs:
runs-on: ubuntu-latest
steps:
- name: Generate image catalogs
uses: cloudnative-pg/postgres-containers/.github/actions/generate-catalogs@main
with:
registry: ghcr.io/cloudnative-pg/postgresql
image-types: minimal,standard
distributions: bookworm,trixie
regex: '(\d+)(?:\.\d+|beta\d+|rc\d+|alpha\d+)-(\d{12})'
output-dir: .
```
This generates:
```
./catalog-minimal-bookworm.yaml
./catalog-standard-bookworm.yaml
./catalog-minimal-trixie.yaml
./catalog-standard-trixie.yaml
```
The generated `kustomization.yaml` will look like:
```
apiVersion: kustomize.config.k8s.io/v1beta1
kind: Kustomization
resources:
- catalog-minimal-bookworm.yaml
- catalog-standard-bookworm.yaml
- catalog-minimal-trixie.yaml
- catalog-standard-trixie.yaml
```

View File

@@ -0,0 +1,74 @@
name: Generate Image Catalogs
description: Generate Image Catalogs
inputs:
registry:
description: "The registry to interrogate"
required: true
image-types:
description: "Image types to retrieve - comma separated values"
required: true
distributions:
description: "OS distributions to retrieve - comma separated values"
required: true
regex:
description: "The regular expression used to retrieve container images"
required: true
output-dir:
description: "The path to output directory"
required: true
family:
description: "The family name to assign to the catalogs"
required: false
runs:
using: composite
steps:
- name: Set up Python
uses: actions/setup-python@e797f83bcb11b83ae66e0230d6156d7c80228e7c # v6
with:
python-version: 3.13
- name: Install Python dependencies
shell: bash
run: |
pip install packaging==25.0 PyYAML==6.0.2
- name: Generate catalogs
shell: bash
env:
REGISTRY: ${{ inputs.registry }}
IMAGE_TYPES: ${{ inputs.image-types }}
DISTRIBUTIONS: ${{ inputs.distributions }}
REGEX: ${{ inputs.regex }}
OUTPUT_DIR: ${{ inputs.output-dir }}
FAMILY: ${{ inputs.family }}
run: |
set -euo pipefail
ARGS=()
if [[ -n "${REGISTRY:-}" ]]; then
ARGS+=( --registry "$REGISTRY" )
fi
if [[ -n "${IMAGE_TYPES:-}" ]]; then
IFS=',' read -r -a image_types <<< "$IMAGE_TYPES"
ARGS+=( --image-types "${image_types[@]}" )
fi
if [[ -n "${DISTRIBUTIONS:-}" ]]; then
IFS=',' read -r -a distributions <<< "$DISTRIBUTIONS"
ARGS+=( --distributions "${distributions[@]}" )
fi
if [[ -n "${REGEX:-}" ]]; then
ARGS+=( --regex "$REGEX" )
fi
if [[ -n "${FAMILY:-}" ]]; then
ARGS+=( --family "$FAMILY" )
fi
ARGS+=( --output-dir "$OUTPUT_DIR" )
echo "Running: python $GITHUB_ACTION_PATH/catalogs_generator.py ${ARGS[*]}"
python "$GITHUB_ACTION_PATH/catalogs_generator.py" "${ARGS[@]}"

View File

@@ -31,11 +31,50 @@ supported_img_types = ["minimal", "standard", "system"]
supported_os_names = ["bullseye", "bookworm", "trixie"]
min_supported_major = 13
repo_name = "cloudnative-pg/postgresql"
full_repo_name = f"ghcr.io/{repo_name}"
pg_regexp = r"(\d+)(?:\.\d+|beta\d+|rc\d+|alpha\d+)-(\d{12})"
default_registry = "ghcr.io/cloudnative-pg/postgresql"
default_family = "postgresql"
default_regex = r"(\d+)(?:\.\d+|beta\d+|rc\d+|alpha\d+)-(\d{12})"
_token_cache = {"value": None, "expires_at": 0}
normalized_pattern = re.compile(
r"""
^(?P<pg_version>\d+(?:\.\d+|beta\d+|rc\d+|alpha\d+)) # A mandatory PostgreSQL version
(?:-(?P<extension_version>\d+(?:\.\d+)+))? # An optional extension version
(?:-(?P<timestamp>\d{12}))? # An optional timestamp
$
""",
re.VERBOSE,
)
# Normalize a tag to make it a valid PEP 440 version.
# Optional capture groups after the Postgres version will
# be appended using a "+"" as a local version segment, and
# concatenated with "." in case there's more then one.
def normalize_tag(tag):
match = normalized_pattern.match(tag)
if not match:
raise ValueError(f"Unrecognized tag format: {tag}")
pg_version = match.group("pg_version")
extension_version = match.group("extension_version")
timestamp = match.group("timestamp")
# Build PEP 440 compliant version
# e.g 17.6, 17.6+202509161052, 17.6+3.6.0.202509161052
extra_match = []
if extension_version:
extra_match.append(extension_version)
if timestamp:
extra_match.append(timestamp)
if extra_match:
normalized_tag = f"{pg_version}+{'.'.join(extra_match)}"
else:
normalized_tag = pg_version
return version.Version(normalized_tag)
def get_json(image_name):
data = check_output(
@@ -52,14 +91,14 @@ def get_json(image_name):
return repo_json
def get_token(repository_name):
def get_token(image_name):
global _token_cache
now = time.time()
if _token_cache["value"] and now < _token_cache["expires_at"]:
return _token_cache["value"]
url = "https://ghcr.io/token?scope=repository:{}:pull".format(repository_name)
url = "https://ghcr.io/token?scope=repository:{}:pull".format(image_name)
with urllib.request.urlopen(url) as response:
data = json.load(response)
token = data["token"]
@@ -70,13 +109,14 @@ def get_token(repository_name):
def get_digest(repository_name, tag):
token = get_token(repository_name)
image_name = repository_name.removeprefix("ghcr.io/")
token = get_token(image_name)
media_types = [
"application/vnd.oci.image.index.v1+json",
"application/vnd.oci.image.manifest.v1+json",
"application/vnd.docker.distribution.manifest.v2+json",
]
url = f"https://ghcr.io/v2/{repository_name}/manifests/{tag}"
url = f"https://ghcr.io/v2/{image_name}/manifests/{tag}"
req = urllib.request.Request(url)
req.add_header("Authorization", "Bearer {}".format(token))
req.add_header("Accept", ",".join(media_types))
@@ -85,6 +125,14 @@ def get_digest(repository_name, tag):
return digest
def get_filename(family, img_type, os_name):
filename_prefix = "catalog"
if family != default_family:
filename_prefix = family
return f"{filename_prefix}-{img_type}-{os_name}.yaml"
def write_catalog(tags, version_re, img_type, os_name, output_dir="."):
image_suffix = f"-{img_type}-{os_name}"
version_re = re.compile(rf"^{version_re}{re.escape(image_suffix)}$")
@@ -97,7 +145,7 @@ def write_catalog(tags, version_re, img_type, os_name, output_dir="."):
tags = [item for item in tags if not exclude_preview.search(item)]
# Sort the tags according to semantic versioning
tags.sort(key=lambda v: version.Version(v.removesuffix(image_suffix)), reverse=True)
tags.sort(key=lambda v: normalize_tag(v.removesuffix(image_suffix)), reverse=True)
results = {}
for item in tags:
@@ -112,16 +160,19 @@ def write_catalog(tags, version_re, img_type, os_name, output_dir="."):
continue
if major not in results:
digest = get_digest(repo_name, item)
results[major] = [f"{full_repo_name}:{item}@{digest}"]
digest = get_digest(args.registry, item)
results[major] = [f"{args.registry}:{item}@{digest}"]
if not results:
raise RuntimeError("No results have been found!")
catalog = {
"apiVersion": "postgresql.cnpg.io/v1",
"kind": "ClusterImageCatalog",
"metadata": {
"name": f"postgresql{image_suffix}",
"name": f"{args.family}{image_suffix}",
"labels": {
"images.cnpg.io/family": "postgresql",
"images.cnpg.io/family": args.family,
"images.cnpg.io/type": img_type,
"images.cnpg.io/os": os_name,
"images.cnpg.io/date": time.strftime("%Y%m%d"),
@@ -137,7 +188,7 @@ def write_catalog(tags, version_re, img_type, os_name, output_dir="."):
}
os.makedirs(output_dir, exist_ok=True)
output_file = os.path.join(output_dir, f"catalog{image_suffix}.yaml")
output_file = os.path.join(output_dir, get_filename(args.family, img_type, os_name))
with open(output_file, "w") as f:
yaml.dump(catalog, f, sort_keys=False)
@@ -146,20 +197,47 @@ if __name__ == "__main__":
parser = argparse.ArgumentParser(
description="CloudNativePG ClusterImageCatalog YAML generator"
)
parser.add_argument(
"--registry",
default=default_registry,
help=f"The registry to interrogate (default: {default_registry})",
)
parser.add_argument(
"--output-dir", default=".", help="Directory to save the YAML files"
)
parser.add_argument(
"--regex",
default=default_regex,
help=f"The regular expression used to retrieve container image. The first capturing group must be the PostgreSQL major version. (default: {default_regex})",
)
parser.add_argument(
"--image-types",
nargs="+",
default=supported_img_types,
help=f"Image types to retrieve (default: {supported_img_types})",
)
parser.add_argument(
"--distributions",
nargs="+",
default=supported_os_names,
help=f"Distributions to retrieve (default: {supported_os_names})",
)
parser.add_argument(
"--family",
default=default_family,
help=f"The family name to assign to the catalogs (default: {default_family})",
)
args = parser.parse_args()
repo_json = get_json(full_repo_name)
repo_json = get_json(args.registry)
tags = repo_json["Tags"]
catalogs = []
for img_type in supported_img_types:
for os_name in supported_os_names:
filename = f"catalog-{img_type}-{os_name}.yaml"
for img_type in args.image_types:
for os_name in args.distributions:
filename = get_filename(args.family, img_type, os_name)
print(f"Generating {filename}")
write_catalog(tags, pg_regexp, img_type, os_name, args.output_dir)
write_catalog(tags, args.regex, img_type, os_name, args.output_dir)
catalogs.append(filename)
kustomization = {

View File

@@ -15,6 +15,7 @@ jobs:
update-catalogs:
runs-on: ubuntu-24.04
steps:
# TODO: remove this step once system images are EOL
- name: Checkout code
uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5
with:
@@ -29,18 +30,15 @@ jobs:
token: ${{ secrets.REPO_GHA_PAT }}
ref: main
- name: Set up Python
uses: actions/setup-python@e797f83bcb11b83ae66e0230d6156d7c80228e7c # v6
with:
python-version: 3.13
- name: Install Python dependencies
run: |
pip install packaging==25.0 PyYAML==6.0.2
- name: Generate catalogs
run: |
python postgres-containers/.github/catalogs_generator.py --output-dir artifacts/image-catalogs/
uses: ./postgres-containers/.github/actions/generate-catalogs
with:
output-dir: artifacts/image-catalogs/
registry: ghcr.io/cloudnative-pg/postgresql
family: postgresql
distributions: bullseye,bookworm,trixie
image-types: minimal,standard,system
regex: '(\d+)(?:\.\d+|beta\d+|rc\d+|alpha\d+)-(\d{12})'
# TODO: remove this step once system images are EOL
- name: Update legacy catalogs