fix: add a grace period before GC deletes orphaned blobs

FindOrphanedBlobs returned any unreferenced blob, so a concurrent dedup
upload (which inserts the blob row before its artifact/local_files row)
could have its S3 object deleted mid-flight. Restrict collection to blobs
older than a 1h grace period.

Refs #71

.woodpecker/docker.yaml

@@ -8,6 +8,8 @@ steps:
     settings:
       registry: git.unkin.net
       repo: git.unkin.net/unkin/artifactapi
+      build_args:
+        VERSION: ${CI_COMMIT_TAG}
       username: droneci
       password:
         from_secret: DRONECI_PASSWORD

Dockerfile

@@ -9,7 +9,8 @@ RUN go mod download
 
 COPY . .
 
-RUN CGO_ENABLED=0 go build -ldflags="-s -w" -o artifactapi ./cmd/artifactapi
+ARG VERSION=dev
+RUN CGO_ENABLED=0 go build -ldflags="-s -w -X main.version=${VERSION}" -o artifactapi ./cmd/artifactapi
 
 FROM gcr.io/distroless/static-debian12:nonroot
 

Makefile

@@ -12,7 +12,7 @@ check-go:
 	fi
 
 build: check-go tidy
-	go build -ldflags="-s -w" -o $(BINARY) ./cmd/artifactapi
+	go build -ldflags="-s -w -X main.version=$(VERSION)" -o $(BINARY) ./cmd/artifactapi
 
 test: check-go
 	go test -race -count=1 ./pkg/... ./internal/...

cmd/artifactapi/main.go

@@ -13,6 +13,8 @@ import (
 	"git.unkin.net/unkin/artifactapi/internal/tui"
 )
 
+var version = "dev"
+
 func main() {
 	if len(os.Args) > 1 && os.Args[1] == "tui" {
 		endpoint := os.Getenv("ARTIFACTAPI_ENDPOINT")
@@ -42,7 +44,7 @@ func main() {
 	ctx, cancel := signal.NotifyContext(context.Background(), syscall.SIGINT, syscall.SIGTERM)
 	defer cancel()
 
-	srv, err := server.New(cfg)
+	srv, err := server.New(cfg, version)
 	if err != nil {
 		slog.Error("failed to create server", "error", err)
 		os.Exit(1)

internal/api/v1/proxy.go

@@ -67,7 +67,7 @@ func (h *ProxyHandler) handleProxy(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
-	result, err := h.engine.Fetch(r.Context(), *remote, path, prov)
+	result, err := h.engine.Fetch(r.Context(), *remote, path, prov, r.Header)
 	if err != nil {
 		var proxyErr *proxy.ProxyError
 		if errors.As(err, &proxyErr) {

internal/database/artifacts.go

@@ -109,16 +109,22 @@ func (db *DB) InsertAccessLog(ctx context.Context, remoteName, path string, cach
 	return err
 }
 
-func (db *DB) FindOrphanedBlobs(ctx context.Context) ([]models.Blob, error) {
+// FindOrphanedBlobs returns blobs no longer referenced by any artifact or
+// local file, restricted to those created before now()-minAge. The age cutoff
+// is a grace period that avoids a TOCTOU race with in-flight dedup uploads,
+// which insert the blob row before the referencing artifact/local_files row.
+func (db *DB) FindOrphanedBlobs(ctx context.Context, minAge time.Duration) ([]models.Blob, error) {
+	cutoff := time.Now().Add(-minAge)
 	rows, err := db.Pool.Query(ctx, `
 		SELECT b.content_hash, b.s3_key, b.size_bytes, b.content_type, b.created_at
 		FROM blobs b
-		WHERE b.content_hash NOT IN (
+		WHERE b.created_at < $1
+		  AND b.content_hash NOT IN (
 			SELECT content_hash FROM artifacts
 			UNION
 			SELECT content_hash FROM local_files
 		)
-	`)
+	`, cutoff)
 	if err != nil {
 		return nil, err
 	}

internal/database/stats.go

@@ -30,6 +30,15 @@ func (db *DB) GetOverviewStats(ctx context.Context) (*models.OverviewStats, erro
 		return nil, err
 	}
 
+	err = db.Pool.QueryRow(ctx, `
+		SELECT COALESCE(SUM(size_bytes), 0)
+		FROM access_log
+		WHERE cache_hit = TRUE AND created_at > NOW() - INTERVAL '30 days'
+	`).Scan(&stats.BandwidthSaved30d)
+	if err != nil {
+		return nil, err
+	}
+
 	return &stats, nil
 }
 

internal/gc/gc.go

@@ -9,6 +9,11 @@ import (
 	"git.unkin.net/unkin/artifactapi/internal/storage"
 )
 
+// blobGracePeriod is how old an orphaned blob must be before GC will delete
+// it. This avoids racing in-flight dedup uploads that insert the blob row
+// before the referencing artifact/local_files row exists.
+const blobGracePeriod = 1 * time.Hour
+
 type Collector struct {
 	db       *database.DB
 	store    *storage.S3
@@ -38,7 +43,7 @@ func (c *Collector) Run(ctx context.Context) {
 func (c *Collector) sweep(ctx context.Context) {
 	start := time.Now()
 
-	orphaned, err := c.db.FindOrphanedBlobs(ctx)
+	orphaned, err := c.db.FindOrphanedBlobs(ctx, blobGracePeriod)
 	if err != nil {
 		slog.Error("gc: find orphaned blobs", "error", err)
 		return

internal/proxy/engine.go

@@ -44,7 +44,7 @@ type FetchResult struct {
 	Source      string // "cache" or "remote"
 }
 
-func (e *Engine) Fetch(ctx context.Context, remote models.Remote, path string, prov provider.Provider) (*FetchResult, error) {
+func (e *Engine) Fetch(ctx context.Context, remote models.Remote, path string, prov provider.Provider, clientHeaders ...http.Header) (*FetchResult, error) {
 	classifier := NewClassifier(prov)
 	class := classifier.Classify(remote, path)
 
@@ -105,8 +105,13 @@ func (e *Engine) Fetch(ctx context.Context, remote models.Remote, path string, p
 		}
 	}
 
+	var fwdHeaders http.Header
+	if len(clientHeaders) > 0 && clientHeaders[0] != nil {
+		fwdHeaders = clientHeaders[0]
+	}
+
 	start := time.Now()
-	result, err := e.fetchFromUpstream(ctx, remote, path, prov, class, ttl)
+	result, err := e.fetchFromUpstream(ctx, remote, path, prov, class, ttl, fwdHeaders)
 	upstreamMS := int(time.Since(start).Milliseconds())
 	if err != nil {
 		if remote.StaleOnError && isNetworkError(err) {
@@ -126,7 +131,7 @@ func (e *Engine) Fetch(ctx context.Context, remote models.Remote, path string, p
 	return result, nil
 }
 
-func (e *Engine) fetchFromUpstream(ctx context.Context, remote models.Remote, path string, prov provider.Provider, class Classification, ttl time.Duration) (*FetchResult, error) {
+func (e *Engine) fetchFromUpstream(ctx context.Context, remote models.Remote, path string, prov provider.Provider, class Classification, ttl time.Duration, clientHeaders http.Header) (*FetchResult, error) {
 	url := prov.UpstreamURL(remote, path)
 
 	authHeaders, err := prov.AuthHeaders(ctx, remote)
@@ -143,6 +148,11 @@ func (e *Engine) fetchFromUpstream(ctx context.Context, remote models.Remote, pa
 			req.Header.Add(k, v)
 		}
 	}
+	if clientHeaders != nil {
+		if accept := clientHeaders.Get("Accept"); accept != "" {
+			req.Header.Set("Accept", accept)
+		}
+	}
 
 	resp, err := http.DefaultClient.Do(req)
 	if err != nil {
@@ -155,6 +165,11 @@ func (e *Engine) fetchFromUpstream(ctx context.Context, remote models.Remote, pa
 		if err == nil && token != "" {
 			req2, _ := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
 			req2.Header.Set("Authorization", "Bearer "+token)
+			if clientHeaders != nil {
+				if accept := clientHeaders.Get("Accept"); accept != "" {
+					req2.Header.Set("Accept", accept)
+				}
+			}
 			resp, err = http.DefaultClient.Do(req2)
 			if err != nil {
 				return nil, &UpstreamError{Err: err}
@@ -184,7 +199,7 @@ func (e *Engine) fetchFromUpstream(ctx context.Context, remote models.Remote, pa
 	}
 
 	contentType := prov.ContentType(path)
-	if ct := resp.Header.Get("Content-Type"); ct != "" && contentType == "application/octet-stream" {
+	if ct := resp.Header.Get("Content-Type"); ct != "" {
 		contentType = ct
 	}
 

internal/server/server.go

@@ -35,6 +35,7 @@ import (
 
 type Server struct {
 	cfg          *config.Config
+	version      string
 	router       chi.Router
 	db           *database.DB
 	cache        *cache.Redis
@@ -45,7 +46,7 @@ type Server struct {
 	gc           *gc.Collector
 }
 
-func New(cfg *config.Config) (*Server, error) {
+func New(cfg *config.Config, version string) (*Server, error) {
 	db, err := database.New(cfg.DatabaseDSN())
 	if err != nil {
 		return nil, fmt.Errorf("database: %w", err)
@@ -68,6 +69,7 @@ func New(cfg *config.Config) (*Server, error) {
 
 	s := &Server{
 		cfg:          cfg,
+		version:      version,
 		db:           db,
 		cache:        redis,
 		store:        s3,
@@ -138,7 +140,7 @@ func (s *Server) handleHealth(w http.ResponseWriter, r *http.Request) {
 func (s *Server) handleRoot(w http.ResponseWriter, r *http.Request) {
 	w.Header().Set("Content-Type", "application/json")
 	w.WriteHeader(http.StatusOK)
-	fmt.Fprint(w, `{"name":"artifactapi","version":"3.0.0-dev"}`)
+	fmt.Fprintf(w, `{"name":"artifactapi","version":"%s"}`, s.version)
 }
 
 func (s *Server) newHTTPServer() *http.Server {

internal/virtual/helm_merger.go

@@ -65,11 +65,12 @@ func (m *HelmMerger) MergeIndexes(members []MemberIndex, proxyBaseURL string) ([
 							if baseHost != "" && extractHost(u) != baseHost {
 								continue
 							}
+							relPath := extractPathRelativeToBase(u, member.BaseURL)
 							ver.URLs[i] = fmt.Sprintf("%s/api/v1/%s/%s/%s",
 								strings.TrimRight(proxyBaseURL, "/"),
 								routePrefix,
 								member.RemoteName,
-								extractPath(u))
+								relPath)
 						} else {
 							ver.URLs[i] = fmt.Sprintf("%s/api/v1/%s/%s/%s",
 								strings.TrimRight(proxyBaseURL, "/"),
@@ -102,6 +103,18 @@ func extractHost(rawURL string) string {
 	return rest[:slashIdx]
 }
 
+func extractPathRelativeToBase(rawURL, baseURL string) string {
+	fullPath := extractPath(rawURL)
+	basePath := extractPath(baseURL)
+	if basePath != "" {
+		basePath = strings.TrimRight(basePath, "/") + "/"
+		if strings.HasPrefix(fullPath, basePath) {
+			return fullPath[len(basePath):]
+		}
+	}
+	return fullPath
+}
+
 func extractPath(rawURL string) string {
 	idx := strings.Index(rawURL, "://")
 	if idx == -1 {

ui/nginx.conf

@@ -5,7 +5,12 @@ server {
     root /usr/share/nginx/html;
     index index.html;
 
-    location ${BASE_PATH} {
+    location ${BASE_PATH}/ {
+        rewrite ^${BASE_PATH}(/.*)$ $1 break;
         try_files $uri $uri/ /index.html;
     }
+
+    location = ${BASE_PATH} {
+        return 301 ${BASE_PATH}/;
+    }
 }

ui/src/pages/Dashboard.tsx

@@ -50,6 +50,11 @@ export function Dashboard() {
           value={formatNumber(stats.total_blobs_deduped)}
           sub="shared blobs"
         />
+        <StatsCard
+          label="Bandwidth Saved"
+          value={formatBytes(stats.bandwidth_saved_30d)}
+          sub="last 30 days"
+        />
       </div>
 
       {health && (