unkinben e24c35f534 feat: background syncer for github_rpm remotes (#108)
## Why

Builds on #107 (merged), which derives `github_rpm` RPM metadata lazily on the
client request path, single-flighted per replica. Two problems remain: the
derive still happens per replica, so across a multi-replica deployment the same
releases are scanned and re-derived N times, multiplying GitHub queries; and a
cold cache blocks the first request on a full derive. GitHub's rate limits are
low (~60/hr unauthenticated, ~5000/hr authenticated), so this needs a single
coordinated syncer with a shared rate limit and conditional requests.

## How

- Add a single per-process background syncer (started at boot, cleanly stopped
  on shutdown) that owns a deduped/coalescing work queue, a worker pool, and one
  global token-bucket rate limiter (`golang.org/x/time/rate`) bound onto the
  github provider so every GitHub call (releases list + each ranged asset GET)
  acquires a token first.
- Re-check each `github_rpm` remote for new/changed releases on its existing
  `mutable_ttl` cadence; derive only new/changed assets incrementally and prune
  assets that disappear upstream. Repodata is served from primed DB rows.
- Prime metadata in the background on remote creation; the create call returns
  immediately.
- Send the stored releases-list `ETag` as `If-None-Match`; a `304` derives
  nothing and is not counted against GitHub's rate limit, so an unchanged repo
  is nearly free.
- Coordinate replicas through a `github_rpm_sync_state` row (`last_synced_at`,
  `etag`, `sync_lease_owner`, `sync_lease_expires`): a periodic scan runs only
  for the replica that atomically claims the lease, bounding total GitHub load
  to ~once per `mutable_ttl` regardless of replica count; the ETag is shared
  through the same row.
- Keep the request path fast: serve current cache, enqueue a prime on an empty
  cache, and return a bounded wait then a retryable `503` rather than blocking
  on a cold derive.
- Add `GITHUB_SYNC_RATE` / `GITHUB_SYNC_BURST` / `GITHUB_SYNC_WORKERS` /
  `GITHUB_SYNC_POLL_INTERVAL` config with conservative defaults (1 req/s, burst
  5, 3 workers, 60s tick) and document the syncer in the README.

## Tests

- Unit (httptest, Range/ETag-aware fixture): `304` releases response derives
  nothing; incremental derive fetches only the newly added asset; the shared
  limiter caps request rate; work-queue enqueues coalesce to one job; prime
  enqueues a job; a held lease stops a second replica from scanning; cold-start
  serves `503` while warm cache serves `200`.
- DB integration (testcontainers postgres): the real lease SQL — one holder at a
  time, recency gate blocks a too-soon periodic re-claim, prime (freshness 0)
  bypasses recency but respects a live lease.
- Docker e2e re-run: `dnf install dotvault` works; prime-on-create derives in the
  background at ~1 req/s (global limiter); `dnf makecache` served fast from the
  priming cache (no cold block); clean shutdown mid-scan, no panics.

## Notes

- Reuses `mutable_ttl` as the check interval (no new per-remote field), per brief.

Reviewed-on: #108
Co-authored-by: Ben Vincent <ben@unkin.net>
Co-committed-by: Ben Vincent <ben@unkin.net>
2026-08-10 21:31:24 +10:00
2026-06-07 19:30:35 +10:00

ArtifactAPI

Caching proxy for package repositories. Single Go binary, 10 package types, content-addressable storage, managed by Terraform.

Quick Start

# Start backing services
docker compose up -d postgres redis minio

# Build and run
make build
./bin/artifactapi

# Frontend (separate container or dev server)
cd ui && npm install && npm run dev

API: http://localhost:8000 | Frontend: http://localhost:5173

Package Types

Type Mutable (auto-detected) Immutable (auto-detected)
generic nothing everything
docker tag manifests, /tags/list blobs, digest manifests
helm index.yaml .tgz charts
pypi simple/* index pages .whl, .tar.gz
npm package metadata .tgz tarballs
rpm repomd.xml, repodata/* .rpm
alpine APKINDEX.tar.gz .apk
puppet v3/modules/*, v3/releases* .tar.gz
terraform */versions */download/*/*
goproxy @v/list, @latest .info, .mod, .zip
github_rpm repodata/* (synthesized) .rpm (redirected)

Providers classify paths automatically. Users only configure what to proxy and TTLs.

github_rpm — GitHub releases as a yum repo (metadata-only, no precache)

A github_rpm remote turns a GitHub repo's releases into a real dnf/yum repository without ever caching the packages. It scans releases for .rpm assets, derives each package's metadata (NEVRA, requires/provides/conflicts/ obsoletes, files, checksum) and synthesizes repodata/ on the fly. Package metadata comes from a ranged GET of just the RPM header (the header sits at the front of the file, so the whole package is never downloaded); the sha256 checksum comes from the GitHub asset digest when present, else a one-time lazy stream. Derived metadata is cached (keyed by asset) so repodata generation is served from primed DB rows, never a cold on-demand derive.

Each package's <location> points back at the remote, which 302-redirects the download to the releases_remote — an existing generic github.com remote that streams the actual bytes. dnf follows the redirect transparently.

Background syncer

A single process-wide background syncer keeps every github_rpm remote's derived metadata current off the client request path:

  • Prime on create. Creating a github_rpm remote enqueues a background prime scan, so its metadata is derived right away without blocking the create call. The first dnf request is served from cache. If a request arrives before the prime lands, it returns a retryable 503 (with Retry-After) rather than serving an empty repo or blocking on a multi-minute derive.
  • Periodic re-check, driven by mutable_ttl. Each remote is re-checked for new or changed releases no more often than its mutable_ttl. New/changed assets are derived incrementally; assets already cached are never re-fetched, and assets that disappear upstream are pruned.
  • ETag / 304 conditional requests. The releases-list ETag is stored per remote and sent as If-None-Match; a 304 Not Modified means nothing changed and the syncer derives nothing. GitHub does not count 304 conditional responses against the rate limit, so an unchanged repo is nearly free — this is the main lever keeping GitHub traffic low.
  • Global rate limit. Every GitHub call (releases list + each ranged asset header GET) passes through a single token-bucket limiter shared across all remotes, so GitHub is never hammered. Configure a token (password) on the remote for the higher authenticated rate limit (~5000/hr vs ~60/hr unauthenticated).
  • Multi-replica coordination. State is shared through the database. Before a periodic scan a replica must atomically claim a per-remote lease (github_rpm_sync_state: last_synced_at, etag, sync_lease_owner, sync_lease_expires); only the winner scans. This bounds total GitHub load to ~once per mutable_ttl regardless of replica count, and the shared etag lets any replica issue the conditional request.
# Backend that serves the actual .rpm bytes from github.com.
resource "artifactapi_remote_generic" "github" {
  name     = "github"
  base_url = "https://github.com"
  patterns = [
    "acme/tools/releases/download/.*\\.rpm$", # allowlist the repo's release assets
  ]
}

resource "artifactapi_remote_github_rpm" "acme-tools" {
  name            = "acme-tools"
  base_url        = "https://api.github.com/repos/acme/tools" # the releases API root
  releases_remote = "github"                                  # backend for downloads
  mutable_ttl     = 3600                                      # release re-scan interval

  # Optional: restrict which release assets become packages (regex on filename).
  patterns = [".*\\.x86_64\\.rpm$", ".*\\.noarch\\.rpm$"]

  # Optional: a token for private repos / higher API rate limits.
  # password = "ghp_..."
}

dnf config: baseurl=https://artifactapi.example/api/v1/remote/acme-tools. The repo is multi-arch (no $basearch needed) — dnf selects matching packages from the synthesized metadata.

Terraform

Remotes and virtuals are managed by Terraform. Each package type has its own resource:

resource "artifactapi_remote_generic" "github" {
  name     = "github"
  base_url = "https://github.com"

  immutable_ttl = 0
  mutable_ttl   = 7200

  patterns = [
    "ducaale/xh/.*/xh-.*-x86_64-unknown-linux-musl.tar.gz$",
    "mikefarah/yq/.*/yq_linux_amd64$",
  ]

  mutable_patterns = [
    ".*/archive/refs/heads/.*\\.tar\\.gz$",
  ]
}

resource "artifactapi_remote_docker" "dockerhub" {
  name     = "dockerhub"
  base_url = "https://registry-1.docker.io"

  immutable_ttl    = 0
  mutable_ttl      = 300
  ban_tags_enabled = true
  ban_tags         = ["latest"]

  patterns = [
    "^library/postgres",
    "^library/redis",
  ]
}

resource "artifactapi_remote_helm" "jetstack" {
  name     = "jetstack"
  base_url = "https://charts.jetstack.io"

  immutable_ttl = 0
  mutable_ttl   = 3600
}

resource "artifactapi_virtual" "helm" {
  name         = "helm"
  package_type = "helm"
  members      = [artifactapi_remote_helm.jetstack.name]
}

Provider: terraform-provider-artifactapi

Serving providers as a registry

A local terraform repo is a real provider registry: upload terraform-provider-{type}_{version}_{os}_{arch}.zip files under {namespace}/{type}/, and Terraform installs them from a bare source address — no .terraformrc mirror config:

terraform {
  required_providers {
    artifactapi = {
      source  = "artifactapi.k8s.syd1.au.unkin.net/<repo>/<type>"
      version = "0.1.2"
    }
  }
}

The Terraform namespace segment is the artifactapi repo name; the provider is matched by type. The registry serves service discovery (/.well-known/terraform.json), the providers.v1 version/download endpoints, and a GPG-signed SHA256SUMS per the provider registry protocol.

Signing needs a GPG key. By default artifactapi generates one on first start and stores it in the database (signing_keys table), so every replica shares it and there's nothing to provision. To bring your own key instead, point TF_SIGNING_KEY_PATH at an armored private key (optionally TF_SIGNING_KEY_PASSPHRASE), which takes precedence over the generated one. TF_PROVIDER_PROTOCOLS (default 5.0,6.0) sets the advertised plugin protocols.

Local docker registry

A local docker repo is a real container registry, not a mirror: it serves the Docker Registry HTTP API V2 for both push and pull, so any client (docker, podman, skopeo, buildah) can use it directly.

docker tag myapp:latest artifactapi.k8s.syd1.au.unkin.net/docker-internal/myapp:latest
docker push        artifactapi.k8s.syd1.au.unkin.net/docker-internal/myapp:latest
docker pull        artifactapi.k8s.syd1.au.unkin.net/docker-internal/myapp:latest

The first path segment after /v2/ is the artifactapi repo name; the remainder is the image name. Blobs and manifests are stored through the shared content-addressable store (deduplicated by digest, reaped by GC once unreferenced); tags are mutable references and re-pushing a tag moves it. Blob uploads support both the monolithic and chunked (POST/PATCH/PUT) flows.

Access Control

Field Default Behaviour
patterns empty (proxy all) If set, only matching paths are proxied. Acts as allowlist.
blocklist empty Matching paths always denied. Checked first.
mutable_patterns empty Override: force paths to mutable TTL.
immutable_patterns empty Override: force paths to immutable TTL.

No patterns + no blocklist = open proxy. Provider handles mutability classification automatically.

API

Proxy (v1)

GET /api/v1/remote/{name}/{path}     Proxy/cache artifact
GET /api/v1/virtual/{name}/{path}    Virtual repo (merged index)
GET /v2/{name}/{path}                Docker Registry v2

Management (v2)

GET/POST        /api/v2/remotes              List / create remotes
GET/PUT/DELETE  /api/v2/remotes/{name}       Read / update / delete remote
GET/DELETE      /api/v2/remotes/{name}/objects  Browse / evict cached objects
GET             /api/v2/stats                Overview stats
GET             /api/v2/health               Service health
POST            /api/v2/probe                Test a remote (fetch without streaming to client)
GET             /api/v2/events               SSE event stream

Architecture

PostgreSQL  ─── config (remotes, virtuals), artifact metadata, access log
Redis       ─── TTL keys, fetch locks, circuit breaker state
S3/MinIO    ─── content-addressable blob storage (blobs/sha256/{hash})

S3 client supports MinIO, Ceph RGW, and AWS S3 (via minio-go).

Environment Variables

Variable Default Description
LISTEN_ADDR :8000 Server listen address
DBHOST localhost PostgreSQL host
DBPORT 5432 PostgreSQL port
DBUSER artifacts PostgreSQL user
DBPASS PostgreSQL password
DBNAME artifacts PostgreSQL database
REDIS_URL redis://localhost:6379 Redis URL
MINIO_ENDPOINT localhost:9000 S3 endpoint
MINIO_ACCESS_KEY S3 access key
MINIO_SECRET_KEY S3 secret key
MINIO_BUCKET artifacts S3 bucket
MINIO_SECURE false Use HTTPS for S3
MINIO_REGION S3 region (AWS)
GITHUB_SYNC_RATE 1 github_rpm syncer global GitHub request rate (req/s), shared across all remotes. 1/s = 3600/hr, under an authenticated token's ~5000/hr; unauthenticated (~60/hr) relies on ETag/304
GITHUB_SYNC_BURST 5 Token-bucket burst for the shared limiter
GITHUB_SYNC_WORKERS 3 Concurrent github_rpm scan workers
GITHUB_SYNC_POLL_INTERVAL 60 Base scheduler tick in seconds; per-remote cadence is its mutable_ttl, enforced by the DB lease

Development

make build       # Build binary
make test        # Unit tests
make e2e         # E2E tests (needs Docker)
make lint        # golangci-lint + go vet
make fmt         # gofmt + goimports

TUI

./bin/artifactapi tui --endpoint http://localhost:8000
S
Description
My terrible vibe coded artifact cache
Readme 2.5 MiB
Languages
Go 89.9%
TypeScript 7.6%
CSS 1.8%
Shell 0.3%
Makefile 0.3%