Evict remote objects from every cache layer (#128)

Evicting a remote object only deleted its Postgres row. The S3 index and the Redis TTL/ETag keys survived, so stale mirror metadata such as EPEL repodata kept being served. - evict the artifact row, S3 index and Redis keys via `Engine.Evict` - evict a directory with `<dir>/*`; other wildcards return 400, unknown remotes 404 - wait on the per-path fetch lock for single-path evicts; return 503 on timeout or a Redis error - wildcard evicts take no lock; a fetch already in flight may re-cache its path - return S3 list errors from `DeletePrefix` instead of deleting nothing Reviewed-on: #128 Co-authored-by: unkin-agent <unkin-agent@unkin.net> Co-committed-by: unkin-agent <unkin-agent@unkin.net>
2026-10-04 22:07:14 +11:00
parent 0b159dad90
commit 6a08539a78
10 changed files with 575 additions and 6 deletions
@@ -16,6 +16,8 @@ import (
 	"sync/atomic"
 	"time"

+	"github.com/jackc/pgx/v5"
+
 	"git.unkin.net/unkin/artifactapi/internal/cache"
 	"git.unkin.net/unkin/artifactapi/internal/database"
 	"git.unkin.net/unkin/artifactapi/internal/provider"
@@ -47,6 +49,8 @@ type Engine struct {
 	// mirror strategy to prefer the mirror currently handling the fewest
 	// requests. Per-replica and approximate, which is fine.
 	inflight sync.Map
+	// evictLockWait bounds how long Evict waits on a held fetch lock.
+	evictLockWait time.Duration
 }

 func NewEngine(db *database.DB, c *cache.Redis, s *storage.S3) *Engine {
@@ -57,6 +61,8 @@ func NewEngine(db *database.DB, c *cache.Redis, s *storage.S3) *Engine {
 		cas:       storage.NewCAS(s),
 		circuit:   NewCircuitBreaker(c),
 		accessLog: make(chan database.AccessLogEntry, accessLogBufferSize),
+
+		evictLockWait: fetchLockTTL,
 	}
 	go e.runAccessLogWriter()
 	return e
@@ -208,6 +214,71 @@ func (e *Engine) Fetch(ctx context.Context, remote models.Remote, path string, p
 	return result, nil
 }

+// Evict drops path from every cache layer (artifact row, index object, Redis
+// freshness and ETag keys) so the next request refetches from upstream. A
+// trailing "/*" evicts every path under that directory.
+func (e *Engine) Evict(ctx context.Context, remoteName, path string) error {
+	prefix, wildcard := strings.CutSuffix(path, "*")
+	if wildcard && !strings.HasSuffix(prefix, "/") {
+		return &ProxyError{Status: http.StatusBadRequest, Message: "wildcard evict must be <dir>/*"}
+	}
+	if _, err := e.db.GetRemote(ctx, remoteName); errors.Is(err, pgx.ErrNoRows) {
+		return &ProxyError{Status: http.StatusNotFound, Message: fmt.Sprintf("remote %q not found", remoteName)}
+	} else if err != nil {
+		return fmt.Errorf("get remote: %w", err)
+	}
+	if !wildcard {
+		if err := e.waitForLock(ctx, remoteName, path); err != nil {
+			return err
+		}
+		defer func() { _ = e.cache.ReleaseLock(context.WithoutCancel(ctx), remoteName, path) }()
+		if err := e.db.DeleteArtifact(ctx, remoteName, path); err != nil {
+			return fmt.Errorf("delete artifact: %w", err)
+		}
+		if err := e.store.Delete(ctx, storage.IndexKey(remoteName, path)); err != nil {
+			return fmt.Errorf("delete index: %w", err)
+		}
+		return e.cache.ForgetPath(ctx, remoteName, path)
+	}
+	// ponytail: no lock for wildcards; a Fetch already in flight under the
+	// prefix can re-cache its path after the evict. Per-path locks over a
+	// directory would close it if that ever matters.
+	if err := e.db.DeleteArtifactsByPrefix(ctx, remoteName, prefix); err != nil {
+		return fmt.Errorf("delete artifacts: %w", err)
+	}
+	if err := e.store.DeletePrefix(ctx, storage.IndexKey(remoteName, prefix)); err != nil {
+		return fmt.Errorf("delete indexes: %w", err)
+	}
+	return e.cache.ForgetPrefix(ctx, remoteName, prefix)
+}
+
+// waitForLock takes the per-path fetch lock so an in-flight Fetch cannot
+// re-set TTL/ETag keys after an evict. It fails with a 503 when the lock
+// cannot be taken within evictLockWait or Redis errors.
+func (e *Engine) waitForLock(ctx context.Context, remoteName, path string) error {
+	deadline := time.Now().Add(e.evictLockWait)
+	for {
+		ok, err := e.cache.AcquireLock(ctx, remoteName, path, fetchLockTTL)
+		if ok {
+			return nil
+		}
+		if ctx.Err() != nil {
+			return ctx.Err()
+		}
+		if err != nil {
+			return &ProxyError{Status: http.StatusServiceUnavailable, Message: fmt.Sprintf("fetch lock: %v", err)}
+		}
+		if time.Now().After(deadline) {
+			return &ProxyError{Status: http.StatusServiceUnavailable, Message: "fetch in progress, retry evict"}
+		}
+		select {
+		case <-ctx.Done():
+			return ctx.Err()
+		case <-time.After(50 * time.Millisecond):
+		}
+	}
+}
+
 // HeadResult carries artifact metadata for a HEAD request. There is no body.
 type HeadResult struct {
 	ContentType string
@@ -0,0 +1,271 @@
+package proxy
+
+import (
+	"context"
+	"errors"
+	"net/http"
+	"net/http/httptest"
+	"sync/atomic"
+	"testing"
+	"time"
+
+	_ "git.unkin.net/unkin/artifactapi/internal/provider/rpm"
+	"git.unkin.net/unkin/artifactapi/internal/storage"
+	"git.unkin.net/unkin/artifactapi/pkg/models"
+)
+
+// changingUpstream serves every path with the current revision, as a mirror
+// does after a sync replaces its repodata. It answers 304 to a matching
+// If-None-Match and counts conditional requests.
+func changingUpstream(t *testing.T) (*httptest.Server, *atomic.Value, *atomic.Int32) {
+	t.Helper()
+	var rev atomic.Value
+	var conditional atomic.Int32
+	rev.Store("rev1")
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		v := rev.Load().(string)
+		etag := `"` + v + `"`
+		w.Header().Set("ETag", etag)
+		if inm := r.Header.Get("If-None-Match"); inm != "" {
+			conditional.Add(1)
+			if inm == etag {
+				w.WriteHeader(http.StatusNotModified)
+				return
+			}
+		}
+		_, _ = w.Write([]byte(v + ":" + r.URL.Path))
+	}))
+	t.Cleanup(srv.Close)
+	return srv, &rev, &conditional
+}
+
+func fetchBody(t *testing.T, r models.Remote, path string) string {
+	t.Helper()
+	res, err := testEngine.Fetch(context.Background(), r, path, prov(t, models.PackageRPM))
+	if err != nil {
+		t.Fatalf("fetch %s: %v", path, err)
+	}
+	return readAll(t, res)
+}
+
+func rpmRemote(t *testing.T, name, baseURL string) models.Remote {
+	return seed(t, models.Remote{Name: name, PackageType: models.PackageRPM, RepoType: models.RepoTypeRemote, BaseURL: baseURL, MutableTTL: 7200, CheckMutable: true})
+}
+
+// cached reports which cache layers hold path: artifact row, index object,
+// Redis TTL key, Redis ETag key.
+type cached struct{ row, index, ttl, etag bool }
+
+// Mutable indexes live in the S3 index; immutable blobs get an artifact row.
+var (
+	indexCached = cached{index: true, ttl: true, etag: true}
+	blobCached  = cached{row: true, ttl: true, etag: true}
+)
+
+func layers(t *testing.T, remote, path string) cached {
+	t.Helper()
+	ctx := context.Background()
+	var c cached
+	_, err := testDB.GetArtifact(ctx, remote, path)
+	c.row = err == nil
+	c.index, err = testEngine.store.Exists(ctx, storage.IndexKey(remote, path))
+	if err != nil {
+		t.Fatalf("stat index %s: %v", path, err)
+	}
+	c.ttl, _ = testCache.CheckTTL(ctx, remote, path)
+	etag, _ := testCache.GetETag(ctx, remote, path)
+	c.etag = etag != ""
+	return c
+}
+
+func TestEvictMutableIndexRefetches(t *testing.T) {
+	requireStack(t)
+	srv, rev, conditional := changingUpstream(t)
+	r := rpmRemote(t, "evict-idx", srv.URL)
+	const path = "8/Everything/x86_64/repodata/repomd.xml"
+
+	if got := fetchBody(t, r, path); got != "rev1:/"+path {
+		t.Fatalf("initial fetch = %q", got)
+	}
+	if c := layers(t, r.Name, path); c != indexCached {
+		t.Fatalf("before evict = %+v, want %+v", c, indexCached)
+	}
+	rev.Store("rev2")
+	if got := fetchBody(t, r, path); got != "rev1:/"+path {
+		t.Fatalf("within TTL = %q, want cached rev1", got)
+	}
+	if err := testEngine.Evict(context.Background(), r.Name, path); err != nil {
+		t.Fatalf("evict: %v", err)
+	}
+	if c := layers(t, r.Name, path); c != (cached{}) {
+		t.Fatalf("after evict = %+v, want every layer gone", c)
+	}
+	conditional.Store(0)
+	if got := fetchBody(t, r, path); got != "rev2:/"+path {
+		t.Fatalf("after evict = %q, want rev2", got)
+	}
+	if n := conditional.Load(); n != 0 {
+		t.Errorf("post-evict fetch revalidated with the evicted ETag (%d conditional requests)", n)
+	}
+}
+
+func TestEvictImmutableBlobDropsRow(t *testing.T) {
+	requireStack(t)
+	srv, _, _ := changingUpstream(t)
+	r := rpmRemote(t, "evict-blob", srv.URL)
+	const path = "8/Everything/x86_64/Packages/a/a-1.0-1.el8.x86_64.rpm"
+	fetchBody(t, r, path)
+	if c := layers(t, r.Name, path); c != blobCached {
+		t.Fatalf("before evict = %+v, want %+v", c, blobCached)
+	}
+	if err := testEngine.Evict(context.Background(), r.Name, path); err != nil {
+		t.Fatalf("evict: %v", err)
+	}
+	if c := layers(t, r.Name, path); c != (cached{}) {
+		t.Fatalf("after evict = %+v, want every layer gone", c)
+	}
+}
+
+func TestEvictRejectsNonDirectoryWildcard(t *testing.T) {
+	requireStack(t)
+	srv, _, _ := changingUpstream(t)
+	r := rpmRemote(t, "evict-bare", srv.URL)
+	const path = "8/Everything/x86_64/repodata/repomd.xml"
+	fetchBody(t, r, path)
+
+	for _, bad := range []string{"*", "8*", "8/Every*"} {
+		var pe *ProxyError
+		if err := testEngine.Evict(context.Background(), r.Name, bad); !errors.As(err, &pe) || pe.Status != http.StatusBadRequest {
+			t.Errorf("evict %s = %v, want 400", bad, err)
+		}
+	}
+	if c := layers(t, r.Name, path); c != (indexCached) {
+		t.Errorf("rejected wildcard evicted layers: %+v", c)
+	}
+}
+
+func TestEvictUnknownRemote(t *testing.T) {
+	requireStack(t)
+	var pe *ProxyError
+	if err := testEngine.Evict(context.Background(), "evict-no-such-remote", "a/b"); !errors.As(err, &pe) || pe.Status != http.StatusNotFound {
+		t.Fatalf("evict unknown remote = %v, want 404", err)
+	}
+}
+
+func holdLock(t *testing.T, remote, path string) {
+	t.Helper()
+	ctx := context.Background()
+	if ok, err := testCache.AcquireLock(ctx, remote, path, time.Minute); !ok || err != nil {
+		t.Fatalf("acquire: %v %v", ok, err)
+	}
+	t.Cleanup(func() { _ = testCache.ReleaseLock(ctx, remote, path) })
+}
+
+func TestEvictWaitsForFetchLock(t *testing.T) {
+	requireStack(t)
+	srv, _, _ := changingUpstream(t)
+	r := rpmRemote(t, "evict-lock", srv.URL)
+	ctx := context.Background()
+	const path = "8/Everything/x86_64/repodata/repomd.xml"
+	holdLock(t, r.Name, path)
+	done := make(chan error, 1)
+	go func() { done <- testEngine.Evict(ctx, r.Name, path) }()
+	select {
+	case err := <-done:
+		t.Fatalf("evict returned while fetch lock held: %v", err)
+	case <-time.After(200 * time.Millisecond):
+	}
+	_ = testCache.ReleaseLock(ctx, r.Name, path)
+	if err := <-done; err != nil {
+		t.Fatalf("evict: %v", err)
+	}
+	ok, err := testCache.AcquireLock(ctx, r.Name, path, time.Second)
+	if !ok || err != nil {
+		t.Fatalf("lock still held after evict: %v %v", ok, err)
+	}
+}
+
+func TestEvictCancelledWhileWaitingDeletesNothing(t *testing.T) {
+	requireStack(t)
+	srv, _, _ := changingUpstream(t)
+	r := rpmRemote(t, "evict-cancel", srv.URL)
+	const path = "8/Everything/x86_64/repodata/repomd.xml"
+	fetchBody(t, r, path)
+	holdLock(t, r.Name, path)
+
+	ctx, cancel := context.WithCancel(context.Background())
+	done := make(chan error, 1)
+	go func() { done <- testEngine.Evict(ctx, r.Name, path) }()
+	time.Sleep(100 * time.Millisecond)
+	cancel()
+	select {
+	case err := <-done:
+		if err == nil {
+			t.Fatal("cancelled evict returned nil")
+		}
+	case <-time.After(2 * time.Second):
+		t.Fatal("cancelled evict did not return")
+	}
+	if c := layers(t, r.Name, path); c != (indexCached) {
+		t.Errorf("cancelled evict deleted layers: %+v", c)
+	}
+}
+
+func TestEvictLockTimeoutIs503(t *testing.T) {
+	requireStack(t)
+	srv, _, _ := changingUpstream(t)
+	r := rpmRemote(t, "evict-timeout", srv.URL)
+	const path = "8/Everything/x86_64/repodata/repomd.xml"
+	fetchBody(t, r, path)
+	holdLock(t, r.Name, path)
+	prev := testEngine.evictLockWait
+	testEngine.evictLockWait = 100 * time.Millisecond
+	t.Cleanup(func() { testEngine.evictLockWait = prev })
+
+	var pe *ProxyError
+	if err := testEngine.Evict(context.Background(), r.Name, path); !errors.As(err, &pe) || pe.Status != http.StatusServiceUnavailable {
+		t.Fatalf("evict = %v, want 503", err)
+	}
+	if c := layers(t, r.Name, path); c != (indexCached) {
+		t.Errorf("timed-out evict deleted layers: %+v", c)
+	}
+}
+
+func TestEvictWildcardClearsPrefixOnly(t *testing.T) {
+	requireStack(t)
+	srv, rev, _ := changingUpstream(t)
+	r := rpmRemote(t, "evict-wild", srv.URL)
+	const (
+		repomd = "8/Everything/x86_64/repodata/repomd.xml"
+		rpm    = "8/Everything/x86_64/Packages/a/a-1.0-1.el8.x86_64.rpm"
+		other  = "9/Everything/x86_64/repodata/repomd.xml"
+	)
+	for _, p := range []string{repomd, rpm, other} {
+		fetchBody(t, r, p)
+	}
+	if c := layers(t, r.Name, rpm); c != blobCached {
+		t.Fatalf("%s before evict = %+v, want %+v", rpm, c, blobCached)
+	}
+	rev.Store("rev2")
+
+	if err := testEngine.Evict(context.Background(), r.Name, "8/Everything/x86_64/*"); err != nil {
+		t.Fatalf("evict: %v", err)
+	}
+	for _, p := range []string{repomd, rpm} {
+		if c := layers(t, r.Name, p); c != (cached{}) {
+			t.Errorf("%s after evict = %+v, want every layer gone", p, c)
+		}
+	}
+	if c := layers(t, r.Name, other); c != (indexCached) {
+		t.Errorf("%s outside prefix = %+v, want %+v", other, c, indexCached)
+	}
+	if got := fetchBody(t, r, repomd); got != "rev2:/"+repomd {
+		t.Errorf("index under prefix = %q, want rev2", got)
+	}
+	if got := fetchBody(t, r, rpm); got != "rev2:/"+rpm {
+		t.Errorf("artifact under prefix = %q, want rev2", got)
+	}
+	if got := fetchBody(t, r, other); got != "rev1:/"+other {
+		t.Errorf("path outside prefix = %q, want cached rev1", got)
+	}
+}