perf: increase worker pool queue buffer size 1 -> ~4k

* Previosuly the worker queue pool buffer size was 1 which applies a backpressure on the fast missing blocks producer

* Define our queue size and why we chose this value

cmd/main.go

@@ -27,7 +27,12 @@ import (
 	"github.com/knadh/koanf/v2"
 )
 
-const defaultGracefulShutdownPeriod = time.Second * 30
+const (
+	defaultGracefulShutdownPeriod = time.Second * 30
+
+	// 24 hrs worth of blocks
+	defaultMaxQueueSize = 17_280
+)
 
 var (
 	build = "dev"
@@ -113,8 +118,11 @@ func main() {
 	poolOpts := pool.PoolOpts{
 		Logg:        lo,
 		WorkerCount: ko.Int("core.pool_size"),
+		// Immidiately allow processing of upto 6 hrs of missing blocks
+		BlocksBuffer: defaultMaxQueueSize / 4,
 	}
 	if ko.Int("core.pool_size") <= 0 {
+		// TODO: Benchamrk to determine optimum size
 		poolOpts.WorkerCount = runtime.NumCPU() * 3
 	}
 	workerPool := pool.NewPool(poolOpts)
@@ -155,9 +163,10 @@ func main() {
 	}
 
 	backfiller := backfiller.New(backfiller.BackfillerOpts{
-		DB:    db,
-		Logg:  lo,
-		Queue: queue,
+		MaxQueueSize: defaultMaxQueueSize,
+		DB:           db,
+		Logg:         lo,
+		Queue:        queue,
 	})
 
 	apiServer := &http.Server{

internal/backfiller/backfiller.go

@@ -12,24 +12,23 @@ import (
 
 type (
 	BackfillerOpts struct {
-		DB    db.DB
-		Logg  *slog.Logger
-		Queue *queue.Queue
+		MaxQueueSize int
+		DB           db.DB
+		Logg         *slog.Logger
+		Queue        *queue.Queue
 	}
 
 	backfiller struct {
-		db     db.DB
-		logg   *slog.Logger
-		queue  *queue.Queue
-		stopCh chan struct{}
-		ticker *time.Ticker
+		maxQueueSize int
+		db           db.DB
+		logg         *slog.Logger
+		queue        *queue.Queue
+		stopCh       chan struct{}
+		ticker       *time.Ticker
 	}
 )
 
-const (
-	verifierInterval = 20 * time.Second
-	epochBlocksCount = 17_280
-)
+const verifierInterval = 20 * time.Second
 
 func New(o BackfillerOpts) *backfiller {
 	return &backfiller{
@@ -58,6 +57,9 @@ func (b *backfiller) Start() {
 				if err := b.Run(true); err != nil {
 					b.logg.Error("verifier tick run error", "err", err)
 				}
+				b.logg.Debug("verifier successful run", "queue_size", b.queue.Size())
+			} else {
+				b.logg.Debug("skipping verifier run")
 			}
 		}
 	}
@@ -84,7 +86,7 @@ func (b *backfiller) Run(skipLatest bool) error {
 	missingBlocksCount := missingBlocks.Count()
 
 	if missingBlocksCount > 0 {
-		if missingBlocksCount >= epochBlocksCount {
+		if missingBlocksCount >= uint(b.maxQueueSize) {
 			b.logg.Warn("large number of blocks missing this may result in degraded RPC performance set FORCE_BACKFILL=* to continue", "missing_blocks", missingBlocksCount)
 			_, ok := os.LookupEnv("FORCE_BACKFILL")
 			if !ok {

internal/pool/pool.go

@@ -8,14 +8,15 @@ import (
 )
 
 type PoolOpts struct {
-	Logg        *slog.Logger
-	WorkerCount int
+	Logg         *slog.Logger
+	WorkerCount  int
+	BlocksBuffer int
 }
 
 func NewPool(o PoolOpts) *pond.WorkerPool {
 	return pond.New(
 		o.WorkerCount,
-		1,
+		o.BlocksBuffer,
 		pond.Strategy(pond.Balanced()),
 		pond.PanicHandler(panicHandler(o.Logg)),
 	)

internal/queue/queue.go

@@ -10,6 +10,7 @@ import (
 
 type (
 	QueueOpts struct {
+		QueueSize int
 		Logg      *slog.Logger
 		Processor *processor.Processor
 		Pool      *pond.WorkerPool
@@ -27,7 +28,7 @@ type (
 func New(o QueueOpts) *Queue {
 	return &Queue{
 		logg:        o.Logg,
-		processChan: make(chan uint64, 17_280),
+		processChan: make(chan uint64, o.QueueSize),
 		stopSignal:  make(chan interface{}),
 		processor:   o.Processor,
 		pool:        o.Pool,