Cache failed handle resolutions with exponentially increasing TTLs (#300

)

api/extra.go

@@ -14,6 +14,7 @@ import (
 
 	"github.com/bluesky-social/indigo/did"
 	"github.com/bluesky-social/indigo/xrpc"
+	arc "github.com/hashicorp/golang-lru/arc/v2"
 	logging "github.com/ipfs/go-log"
 	"go.opentelemetry.io/contrib/instrumentation/net/http/otelhttp"
 	otel "go.opentelemetry.io/otel"
@@ -75,8 +76,26 @@ type HandleResolver interface {
 	ResolveHandleToDid(ctx context.Context, handle string) (string, error)
 }
 
+type failCacheItem struct {
+	err       error
+	count     int
+	expiresAt time.Time
+}
+
 type ProdHandleResolver struct {
-	ReqMod func(*http.Request, string) error
+	ReqMod    func(*http.Request, string) error
+	FailCache *arc.ARCCache[string, *failCacheItem]
+}
+
+func NewProdHandleResolver(failureCacheSize int) (*ProdHandleResolver, error) {
+	failureCache, err := arc.NewARC[string, *failCacheItem](failureCacheSize)
+	if err != nil {
+		return nil, err
+	}
+
+	return &ProdHandleResolver{
+		FailCache: failureCache,
+	}, nil
 }
 
 func (dr *ProdHandleResolver) ResolveHandleToDid(ctx context.Context, handle string) (string, error) {
@@ -86,6 +105,18 @@ func (dr *ProdHandleResolver) ResolveHandleToDid(ctx context.Context, handle str
 	ctx, span := otel.Tracer("resolver").Start(ctx, "ResolveHandleToDid")
 	defer span.End()
 
+	var cachedFailureCount int
+
+	if dr.FailCache != nil {
+		if item, ok := dr.FailCache.Get(handle); ok {
+			cachedFailureCount = item.count
+			if item.expiresAt.After(time.Now()) {
+				return "", item.err
+			}
+			dr.FailCache.Remove(handle)
+		}
+	}
+
 	var wkres, dnsres string
 	var wkerr, dnserr error
 
@@ -117,7 +148,28 @@ func (dr *ProdHandleResolver) ResolveHandleToDid(ctx context.Context, handle str
 		return wkres, nil
 	}
 
-	return "", errors.Join(fmt.Errorf("no did record found for handle %q", handle), dnserr, wkerr)
+	err := errors.Join(fmt.Errorf("no did record found for handle %q", handle), dnserr, wkerr)
+
+	if dr.FailCache != nil {
+		cachedFailureCount++
+		expireAt := time.Now().Add(time.Millisecond * 100)
+		if cachedFailureCount > 1 {
+			// exponential backoff
+			expireAt = time.Now().Add(time.Millisecond * 100 * time.Duration(cachedFailureCount*cachedFailureCount))
+			// Clamp to one hour
+			if expireAt.After(time.Now().Add(time.Hour)) {
+				expireAt = time.Now().Add(time.Hour)
+			}
+		}
+
+		dr.FailCache.Add(handle, &failCacheItem{
+			err:       err,
+			expiresAt: expireAt,
+			count:     cachedFailureCount,
+		})
+	}
+
+	return "", err
 }
 
 func (dr *ProdHandleResolver) resolveWellKnown(ctx context.Context, handle string) (string, error) {

cmd/bigsky/main.go

@@ -297,7 +297,10 @@ func Bigsky(cctx *cli.Context) error {
 		blobstore = &blobs.DiskBlobStore{bsdir}
 	}
 
-	prodHR := api.ProdHandleResolver{}
+	prodHR, err := api.NewProdHandleResolver(100_000)
+	if err != nil {
+		return fmt.Errorf("failed to set up handle resolver: %w", err)
+	}
 	if rlskip != "" {
 		prodHR.ReqMod = func(req *http.Request, host string) error {
 			if strings.HasSuffix(host, ".bsky.social") {
@@ -307,7 +310,7 @@ func Bigsky(cctx *cli.Context) error {
 		}
 	}
 
-	var hr api.HandleResolver = &prodHR
+	var hr api.HandleResolver = prodHR
 	if cctx.StringSlice("handle-resolver-hosts") != nil {
 		hr = &api.TestHandleResolver{
 			TrialHosts: cctx.StringSlice("handle-resolver-hosts"),

go.mod

@@ -12,6 +12,7 @@ require (
 	github.com/gorilla/websocket v1.5.0
 	github.com/hashicorp/go-retryablehttp v0.7.2
 	github.com/hashicorp/golang-lru v0.5.4
+	github.com/hashicorp/golang-lru/arc/v2 v2.0.6
 	github.com/hashicorp/golang-lru/v2 v2.0.6
 	github.com/icrowley/fake v0.0.0-20221112152111-d7b7e2276db2
 	github.com/ipfs/go-block-format v0.1.2

go.sum

@@ -235,6 +235,8 @@ github.com/hashicorp/golang-lru v0.5.0/go.mod h1:/m3WP610KZHVQ1SGc6re/UDhFvYD7pJ
 github.com/hashicorp/golang-lru v0.5.1/go.mod h1:/m3WP610KZHVQ1SGc6re/UDhFvYD7pJ4Ao+sR/qLZy8=
 github.com/hashicorp/golang-lru v0.5.4 h1:YDjusn29QI/Das2iO9M0BHnIbxPeyuCHsjMW+lJfyTc=
 github.com/hashicorp/golang-lru v0.5.4/go.mod h1:iADmTwqILo4mZ8BN3D2Q6+9jd8WM5uGBxy+E8yxSoD4=
+github.com/hashicorp/golang-lru/arc/v2 v2.0.6 h1:4NU7uP5vSoK6TbaMj3NtY478TTAWLso/vL1gpNrInHg=
+github.com/hashicorp/golang-lru/arc/v2 v2.0.6/go.mod h1:cfdDIX05DWvYV6/shsxDfa/OVcRieOt+q4FnM8x+Xno=
 github.com/hashicorp/golang-lru/v2 v2.0.6 h1:3xi/Cafd1NaoEnS/yDssIiuVeDVywU0QdFGl3aQaQHM=
 github.com/hashicorp/golang-lru/v2 v2.0.6/go.mod h1:QeFd9opnmA6QUJc5vARoKUSoFhyfM2/ZepoAG6RGpeM=
 github.com/huin/goupnp v1.0.3 h1:N8No57ls+MnjlB+JPiCVSOyy/ot7MJTqlo7rn+NYSqQ=