Merge branch 'develop' into refcell/challenger/types

go.mod

@@ -190,6 +190,6 @@ require (
 	nhooyr.io/websocket v1.8.7 // indirect
 )
 
-replace github.com/ethereum/go-ethereum v1.11.6 => github.com/ethereum-optimism/op-geth v1.101105.2-0.20230502202351-9cc072e922f6
+replace github.com/ethereum/go-ethereum v1.11.6 => github.com/ethereum-optimism/op-geth v1.101105.2-0.20230526154603-bdab05ca786f
 
 //replace github.com/ethereum/go-ethereum v1.11.6 => ../go-ethereum

go.sum

@@ -151,8 +151,8 @@ github.com/envoyproxy/protoc-gen-validate v0.1.0/go.mod h1:iSmxcyjqTsJpI2R4NaDN7
 github.com/etcd-io/bbolt v1.3.3/go.mod h1:ZF2nL25h33cCyBtcyWeZ2/I3HQOfTP+0PIEvHjkjCrw=
 github.com/ethereum-optimism/go-ethereum-hdwallet v0.1.3 h1:RWHKLhCrQThMfch+QJ1Z8veEq5ZO3DfIhZ7xgRP9WTc=
 github.com/ethereum-optimism/go-ethereum-hdwallet v0.1.3/go.mod h1:QziizLAiF0KqyLdNJYD7O5cpDlaFMNZzlxYNcWsJUxs=
-github.com/ethereum-optimism/op-geth v1.101105.2-0.20230502202351-9cc072e922f6 h1:Fh9VBmDCwjVn8amx1Dfrx+hIh16C/FDkS17EN25MGO8=
-github.com/ethereum-optimism/op-geth v1.101105.2-0.20230502202351-9cc072e922f6/go.mod h1:X9t7oeerFMU9/zMIjZKT/jbIca+O05QqtBTLjL+XVeA=
+github.com/ethereum-optimism/op-geth v1.101105.2-0.20230526154603-bdab05ca786f h1:+yZN8K/4AIN5f+gazMwZAeqzDG2EL2GydMrccjnEK+A=
+github.com/ethereum-optimism/op-geth v1.101105.2-0.20230526154603-bdab05ca786f/go.mod h1:X9t7oeerFMU9/zMIjZKT/jbIca+O05QqtBTLjL+XVeA=
 github.com/fasthttp-contrib/websocket v0.0.0-20160511215533-1f3b11f56072/go.mod h1:duJ4Jxv5lDcvg4QuQr0oowTf7dz4/CR8NtyCooz9HL8=
 github.com/fatih/structs v1.1.0/go.mod h1:9NiDSp5zOcgEDl+j00MP/WkGVPOlPRLejGD8Ga6PJ7M=
 github.com/fjl/memsize v0.0.1 h1:+zhkb+dhUgx0/e+M8sF0QqiouvMQUiKR+QYvdxIOKcQ=

indexer/.gitignore

+docker-compose.dev.yml
+.env
+

indexer/README.md

+# @eth-optimism/indexer
+
+## Getting started
+
+### Run indexer vs goerli
+
+- install docker
+- `cp example.env .env`
+- fill in .env
+- run `docker-compose up` to start the indexer vs optimism goerli network
+
+### Run indexer with go
+
+See the flags in `flags.go` for reference of what command line flags to pass to `go run`
+
+### Run indexer vs devnet
+
+TODO add indexer to the optimism devnet compose file (previously removed for breaking CI)
+
+### Run indexer vs a custom configuration
+
+`docker-compose.dev.yml` is git ignored.   Fill in your own docker-compose file here.
+

indexer/docker-compose.yml

+version: '3.8'
+
+services:
+  postgres:
+    image: postgres:latest
+    environment:
+      - POSTGRES_USER=db_username
+      - POSTGRES_PASSWORD=db_password
+      - POSTGRES_DB=db_name
+      - PGDATA=/data/postgres
+      - POSTGRES_HOST_AUTH_METHOD=trust
+    healthcheck:
+      test: [ "CMD-SHELL", "pg_isready -q -U db_username -d db_name" ]
+    ports:
+      - "5434:5432"
+    volumes:
+      - postgres_data:/data/postgres
+
+  indexer:
+    build:
+      context: ..
+      dockerfile: indexer/Dockerfile
+    healthcheck:
+      test: wget localhost:8080/healthz -q -O - > /dev/null 2>&1
+    environment:
+      # Note that you must index goerli with INDEXER_BEDROCK=false first, then
+      # reindex with INDEXER_BEDROCK=true or seed the database
+      - INDEXER_BEDROCK=${INDEXER_BEDROCK_GOERLI:-true}
+      - INDEXER_BUILD_ENV=${INDEXER_BUILD_ENV:-development}
+      - INDEXER_DB_PORT=${INDEXER_DB_PORT:-5432}
+      - INDEXER_DB_USER=${INDEXER_DB_USER:-db_username}
+      - INDEXER_DB_PASSWORD=${INDEXER_DB_PASSWORD:-db_password}
+      - INDEXER_DB_NAME=${INDEXER_DB_NAME:-db_name}
+      - INDEXER_DB_HOST=${INDEXER_DB_HOST:-postgres}
+      - INDEXER_CHAIN_ID=${INDEXER_CHAIN_ID:-5}
+      - INDEXER_L1_ETH_RPC=$INDEXER_L1_ETH_RPC
+      - INDEXER_L2_ETH_RPC=$INDEXER_L2_ETH_RPC
+      - INDEXER_REST_HOSTNAME=0.0.0.0
+      - INDEXER_REST_PORT=8080
+      - INDEXER_BEDROCK_L1_STANDARD_BRIDGE=0
+      - INDEXER_BEDROCK_L1_STANDARD_BRIDGE=0x636Af16bf2f682dD3109e60102b8E1A089FedAa8
+      - INDEXER_BEDROCK_OPTIMISM_PORTAL=0xB7040fd32359688346A3D1395a42114cf8E3b9b2
+    ports:
+      - 8080:8080
+    depends_on:
+      postgres:
+        condition: service_healthy
+        
+volumes:
+  postgres_data:

indexer/example.env

+# Fill me in with goerli uris and run docker-compose up to run indexer vs goerli
+INDEXER_L1_ETH_RPC=FILL_ME_IN
+INDEXER_L2_ETH_RPC=FILL_ME_IN
+

op-e2e/actions/system_config_test.go

@@ -212,7 +212,14 @@ func TestGPOParamsChange(gt *testing.T) {
 	receipt := alice.LastTxReceipt(t)
 	require.Equal(t, basefee, receipt.L1GasPrice, "L1 gas price matches basefee of L1 origin")
 	require.NotZero(t, receipt.L1GasUsed, "L2 tx uses L1 data")
-	l1Cost := types.L1Cost(receipt.L1GasUsed.Uint64(), basefee, big.NewInt(2100), big.NewInt(1000_000))
+	require.Equal(t,
+		new(big.Float).Mul(
+			new(big.Float).SetInt(basefee),
+			new(big.Float).Mul(new(big.Float).SetInt(receipt.L1GasUsed), receipt.FeeScalar),
+		),
+		new(big.Float).SetInt(receipt.L1Fee), "fee field in receipt matches gas used times scalar times basefee")
+	// receipt.L1GasUsed includes the overhead already, so subtract that before passing it into the L1 cost func
+	l1Cost := types.L1Cost(receipt.L1GasUsed.Uint64()-2100, basefee, big.NewInt(2100), big.NewInt(1000_000))
 	require.Equal(t, l1Cost, receipt.L1Fee, "L1 fee is computed with standard GPO params")
 	require.Equal(t, "1", receipt.FeeScalar.String(), "1000_000 divided by 6 decimals = float(1)")
 
@@ -268,7 +275,8 @@ func TestGPOParamsChange(gt *testing.T) {
 	receipt = alice.LastTxReceipt(t)
 	require.Equal(t, basefeeGPOUpdate, receipt.L1GasPrice, "L1 gas price matches basefee of L1 origin")
 	require.NotZero(t, receipt.L1GasUsed, "L2 tx uses L1 data")
-	l1Cost = types.L1Cost(receipt.L1GasUsed.Uint64(), basefeeGPOUpdate, big.NewInt(1000), big.NewInt(2_300_000))
+	// subtract overhead from L1GasUsed receipt field, types.L1Cost applies it again
+	l1Cost = types.L1Cost(receipt.L1GasUsed.Uint64()-1000, basefeeGPOUpdate, big.NewInt(1000), big.NewInt(2_300_000))
 	require.Equal(t, l1Cost, receipt.L1Fee, "L1 fee is computed with updated GPO params")
 	require.Equal(t, "2.3", receipt.FeeScalar.String(), "2_300_000 divided by 6 decimals = float(2.3)")
 
@@ -288,7 +296,8 @@ func TestGPOParamsChange(gt *testing.T) {
 	receipt = alice.LastTxReceipt(t)
 	require.Equal(t, basefee, receipt.L1GasPrice, "L1 gas price matches basefee of L1 origin")
 	require.NotZero(t, receipt.L1GasUsed, "L2 tx uses L1 data")
-	l1Cost = types.L1Cost(receipt.L1GasUsed.Uint64(), basefee, big.NewInt(1000), big.NewInt(2_300_000))
+	// subtract overhead from L1GasUsed receipt field, types.L1Cost applies it again
+	l1Cost = types.L1Cost(receipt.L1GasUsed.Uint64()-1000, basefee, big.NewInt(1000), big.NewInt(2_300_000))
 	require.Equal(t, l1Cost, receipt.L1Fee, "L1 fee is computed with updated GPO params")
 	require.Equal(t, "2.3", receipt.FeeScalar.String(), "2_300_000 divided by 6 decimals = float(2.3)")
 }

op-e2e/op_geth_test.go

@@ -100,6 +100,126 @@ func TestInvalidDepositInFCU(t *testing.T) {
 	require.Equal(t, 0, balance.Cmp(common.Big0))
 }
 
+// TestGethOnlyPendingBlockIsLatest walks through an engine-API block building job,
+// and asserts that the pending block is set to match the latest block at every stage,
+// for stability and tx-privacy.
+func TestGethOnlyPendingBlockIsLatest(t *testing.T) {
+	InitParallel(t)
+	cfg := DefaultSystemConfig(t)
+	cfg.DeployConfig.FundDevAccounts = true
+	ctx, cancel := context.WithTimeout(context.Background(), 60*time.Second)
+	defer cancel()
+	opGeth, err := NewOpGeth(t, ctx, &cfg)
+	require.NoError(t, err)
+	defer opGeth.Close()
+
+	checkPending := func(stage string, number uint64) {
+		// TODO(CLI-4044): pending-block ID change
+		pendingBlock, err := opGeth.L2Client.BlockByNumber(ctx, big.NewInt(-1))
+		require.NoError(t, err, "failed to fetch pending block at stage "+stage)
+		require.Equal(t, number, pendingBlock.NumberU64(), "pending block must have expected number")
+		latestBlock, err := opGeth.L2Client.BlockByNumber(ctx, nil)
+		require.NoError(t, err, "failed to fetch latest block at stage "+stage)
+		require.Equal(t, pendingBlock.Hash(), latestBlock.Hash(), "pending and latest do not match at stage "+stage)
+	}
+
+	checkPending("genesis", 0)
+
+	amount := big.NewInt(42) // send 42 wei
+
+	aliceStartBalance, err := opGeth.L2Client.PendingBalanceAt(ctx, cfg.Secrets.Addresses().Alice)
+	require.NoError(t, err)
+	require.True(t, aliceStartBalance.Cmp(big.NewInt(0)) > 0, "alice must be funded")
+
+	checkPendingBalance := func() {
+		pendingBalance, err := opGeth.L2Client.PendingBalanceAt(ctx, cfg.Secrets.Addresses().Alice)
+		require.NoError(t, err)
+		require.Equal(t, pendingBalance, aliceStartBalance, "pending balance must still be the same")
+	}
+
+	startBlock, err := opGeth.L2Client.BlockByNumber(ctx, nil)
+	require.NoError(t, err)
+
+	signer := types.LatestSigner(opGeth.L2ChainConfig)
+	tip := big.NewInt(7_000_000_000) // 7 gwei tip
+	tx := types.MustSignNewTx(cfg.Secrets.Alice, signer, &types.DynamicFeeTx{
+		ChainID:   big.NewInt(int64(cfg.DeployConfig.L2ChainID)),
+		Nonce:     0,
+		GasTipCap: tip,
+		GasFeeCap: new(big.Int).Add(startBlock.BaseFee(), tip),
+		Gas:       1_000_000,
+		To:        &cfg.Secrets.Addresses().Bob,
+		Value:     amount,
+		Data:      nil,
+	})
+	require.NoError(t, opGeth.L2Client.SendTransaction(ctx, tx), "send tx to make pending work different")
+	checkPending("prepared", 0)
+
+	rpcClient, err := opGeth.node.Attach()
+	require.NoError(t, err)
+	defer rpcClient.Close()
+
+	// Wait for tx to be in tx-pool, for it to be picked up in block building
+	var txPoolStatus struct {
+		Pending hexutil.Uint64 `json:"pending"`
+	}
+	for i := 0; i < 5; i++ {
+		require.NoError(t, rpcClient.CallContext(ctx, &txPoolStatus, "txpool_status"))
+		if txPoolStatus.Pending == 0 {
+			time.Sleep(time.Second)
+		} else {
+			break
+		}
+	}
+	require.NotZero(t, txPoolStatus.Pending, "must have pending tx in pool")
+
+	checkPending("in-pool", 0)
+	checkPendingBalance()
+
+	// start building a block
+	attrs, err := opGeth.CreatePayloadAttributes()
+	require.NoError(t, err)
+	attrs.NoTxPool = false // we want to include a tx
+	fc := eth.ForkchoiceState{
+		HeadBlockHash: opGeth.L2Head.BlockHash,
+		SafeBlockHash: opGeth.L2Head.BlockHash,
+	}
+	res, err := opGeth.l2Engine.ForkchoiceUpdate(ctx, &fc, attrs)
+	require.NoError(t, err)
+
+	checkPending("building", 0)
+	checkPendingBalance()
+
+	// Now we have to wait until the block-building job picks up the tx from the tx-pool.
+	// See go routine that spins up in buildPayload() func in payload_building.go in miner package.
+	// We can't check it, we don't want to finish block-building prematurely, and so we have to wait.
+	time.Sleep(time.Second * 4) // conservatively wait 4 seconds, CI might lag during block building.
+
+	// retrieve the block
+	payload, err := opGeth.l2Engine.GetPayload(ctx, *res.PayloadID)
+	require.NoError(t, err)
+	checkPending("retrieved", 0)
+	require.Len(t, payload.Transactions, 2, "must include L1 info tx and tx from alice")
+	checkPendingBalance()
+
+	// process the block
+	status, err := opGeth.l2Engine.NewPayload(ctx, payload)
+	require.NoError(t, err)
+	require.Equal(t, eth.ExecutionValid, status.Status)
+	checkPending("processed", 0)
+	checkPendingBalance()
+
+	// make the block canonical
+	fc = eth.ForkchoiceState{
+		HeadBlockHash: payload.BlockHash,
+		SafeBlockHash: payload.BlockHash,
+	}
+	res, err = opGeth.l2Engine.ForkchoiceUpdate(ctx, &fc, nil)
+	require.NoError(t, err)
+	require.Equal(t, eth.ExecutionValid, res.PayloadStatus.Status)
+	checkPending("canonical", 1)
+}
+
 func TestPreregolith(t *testing.T) {
 	InitParallel(t)
 	futureTimestamp := hexutil.Uint64(4)

op-e2e/setup.go

@@ -679,7 +679,7 @@ func (sys *System) newMockNetPeer() (host.Host, error) {
 	_ = ps.AddPubKey(p, sk.GetPublic())
 
 	ds := sync.MutexWrap(ds.NewMapDatastore())
-	eps, err := store.NewExtendedPeerstore(context.Background(), log.Root(), clock.SystemClock, ps, ds)
+	eps, err := store.NewExtendedPeerstore(context.Background(), log.Root(), clock.SystemClock, ps, ds, 24*time.Hour)
 	if err != nil {
 		return nil, err
 	}

op-e2e/system_test.go

@@ -241,12 +241,14 @@ func TestPendingGasLimit(t *testing.T) {
 	cfg.GethOptions["sequencer"] = []GethOption{
 		func(ethCfg *ethconfig.Config, nodeCfg *node.Config) error {
 			ethCfg.Miner.GasCeil = 10_000_000
+			ethCfg.Miner.RollupComputePendingBlock = true
 			return nil
 		},
 	}
 	cfg.GethOptions["verifier"] = []GethOption{
 		func(ethCfg *ethconfig.Config, nodeCfg *node.Config) error {
 			ethCfg.Miner.GasCeil = 9_000_000
+			ethCfg.Miner.RollupComputePendingBlock = true
 			return nil
 		},
 	}
@@ -1230,6 +1232,14 @@ func TestFees(t *testing.T) {
 	require.Nil(t, err)
 	require.Equal(t, l1Fee, gpoL1Fee, "l1 fee mismatch")
 
+	require.Equal(t, receipt.L1Fee, l1Fee, "l1 fee in receipt is correct")
+	require.Equal(t,
+		new(big.Float).Mul(
+			new(big.Float).SetInt(l1Header.BaseFee),
+			new(big.Float).Mul(new(big.Float).SetInt(receipt.L1GasUsed), receipt.FeeScalar),
+		),
+		new(big.Float).SetInt(receipt.L1Fee), "fee field in receipt matches gas used times scalar times basefee")
+
 	// Calculate total fee
 	baseFeeRecipientDiff.Add(baseFeeRecipientDiff, coinbaseDiff)
 	totalFee := new(big.Int).Add(baseFeeRecipientDiff, l1FeeRecipientDiff)
@@ -1371,3 +1381,45 @@ func latestBlock(t *testing.T, client *ethclient.Client) uint64 {
 	require.Nil(t, err, "Error getting latest block")
 	return blockAfter
 }
+
+// TestPendingBlockIsLatest tests that we serve the latest block as pending block
+func TestPendingBlockIsLatest(t *testing.T) {
+	InitParallel(t)
+
+	cfg := DefaultSystemConfig(t)
+	sys, err := cfg.Start()
+	require.Nil(t, err, "Error starting up system")
+	defer sys.Close()
+	l2Seq := sys.Clients["sequencer"]
+
+	t.Run("block", func(t *testing.T) {
+		for i := 0; i < 10; i++ {
+			// TODO(CLI-4044): pending-block ID change
+			pending, err := l2Seq.BlockByNumber(context.Background(), big.NewInt(-1))
+			require.NoError(t, err)
+			latest, err := l2Seq.BlockByNumber(context.Background(), nil)
+			require.NoError(t, err)
+			if pending.NumberU64() == latest.NumberU64() {
+				require.Equal(t, pending.Hash(), latest.Hash(), "pending must exactly match latest block")
+				return
+			}
+			// re-try until we have the same number, as the requests are not an atomic bundle, and the sequencer may create a block.
+		}
+		t.Fatal("failed to get pending block with same number as latest block")
+	})
+	t.Run("header", func(t *testing.T) {
+		for i := 0; i < 10; i++ {
+			// TODO(CLI-4044): pending-block ID change
+			pending, err := l2Seq.HeaderByNumber(context.Background(), big.NewInt(-1))
+			require.NoError(t, err)
+			latest, err := l2Seq.HeaderByNumber(context.Background(), nil)
+			require.NoError(t, err)
+			if pending.Number.Uint64() == latest.Number.Uint64() {
+				require.Equal(t, pending.Hash(), latest.Hash(), "pending must exactly match latest header")
+				return
+			}
+			// re-try until we have the same number, as the requests are not an atomic bundle, and the sequencer may create a block.
+		}
+		t.Fatal("failed to get pending header with same number as latest header")
+	})
+}

op-node/p2p/host.go

@@ -141,7 +141,8 @@ func (conf *Config) Host(log log.Logger, reporter metrics.Reporter, metrics Host
 		return nil, fmt.Errorf("failed to open peerstore: %w", err)
 	}
 
-	ps, err := store.NewExtendedPeerstore(context.Background(), log, clock.SystemClock, basePs, conf.Store)
+	peerScoreParams := conf.PeerScoringParams()
+	ps, err := store.NewExtendedPeerstore(context.Background(), log, clock.SystemClock, basePs, conf.Store, peerScoreParams.RetainScore)
 	if err != nil {
 		return nil, fmt.Errorf("failed to open extended peerstore: %w", err)
 	}

op-node/p2p/peer_scores_test.go

@@ -76,7 +76,7 @@ func getNetHosts(testSuite *PeerScoresTestSuite, ctx context.Context, n int) []h
 	log := testlog.Logger(testSuite.T(), log.LvlError)
 	for i := 0; i < n; i++ {
 		swarm := tswarm.GenSwarm(testSuite.T())
-		eps, err := store.NewExtendedPeerstore(ctx, log, clock.SystemClock, swarm.Peerstore(), sync.MutexWrap(ds.NewMapDatastore()))
+		eps, err := store.NewExtendedPeerstore(ctx, log, clock.SystemClock, swarm.Peerstore(), sync.MutexWrap(ds.NewMapDatastore()), 1*time.Hour)
 		netw := &customPeerstoreNetwork{swarm, eps}
 		require.NoError(testSuite.T(), err)
 		h := bhost.NewBlankHost(netw)
@@ -99,7 +99,7 @@ func newGossipSubs(testSuite *PeerScoresTestSuite, ctx context.Context, hosts []
 		dataStore := sync.MutexWrap(ds.NewMapDatastore())
 		peerStore, err := pstoreds.NewPeerstore(context.Background(), dataStore, pstoreds.DefaultOpts())
 		require.NoError(testSuite.T(), err)
-		extPeerStore, err := store.NewExtendedPeerstore(context.Background(), logger, clock.SystemClock, peerStore, dataStore)
+		extPeerStore, err := store.NewExtendedPeerstore(context.Background(), logger, clock.SystemClock, peerStore, dataStore, 1*time.Hour)
 		require.NoError(testSuite.T(), err)
 
 		scorer := NewScorer(

op-node/p2p/store/extended.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"time"
 
 	"github.com/ethereum-optimism/optimism/op-service/clock"
 	"github.com/ethereum/go-ethereum/log"
@@ -19,12 +20,12 @@ type extendedStore struct {
 	*ipBanBook
 }
 
-func NewExtendedPeerstore(ctx context.Context, logger log.Logger, clock clock.Clock, ps peerstore.Peerstore, store ds.Batching) (ExtendedPeerstore, error) {
+func NewExtendedPeerstore(ctx context.Context, logger log.Logger, clock clock.Clock, ps peerstore.Peerstore, store ds.Batching, scoreRetention time.Duration) (ExtendedPeerstore, error) {
 	cab, ok := peerstore.GetCertifiedAddrBook(ps)
 	if !ok {
 		return nil, errors.New("peerstore should also be a certified address book")
 	}
-	sb, err := newScoreBook(ctx, logger, clock, store)
+	sb, err := newScoreBook(ctx, logger, clock, store, scoreRetention)
 	if err != nil {
 		return nil, fmt.Errorf("create scorebook: %w", err)
 	}

op-node/p2p/store/records_book.go

@@ -102,6 +102,9 @@ func (d *recordsBook[K, V]) deleteRecord(key K) error {
 
 func (d *recordsBook[K, V]) getRecord(key K) (v V, err error) {
 	if val, ok := d.cache.Get(key); ok {
+		if d.hasExpired(val) {
+			return v, UnknownRecordErr
+		}
 		return val, nil
 	}
 	data, err := d.store.Get(d.ctx, d.dsKey(key))
@@ -114,6 +117,9 @@ func (d *recordsBook[K, V]) getRecord(key K) (v V, err error) {
 	if err := v.UnmarshalBinary(data); err != nil {
 		return v, fmt.Errorf("invalid value for key %v: %w", key, err)
 	}
+	if d.hasExpired(v) {
+		return v, UnknownRecordErr
+	}
 	d.cache.Add(key, v)
 	return v, nil
 }
@@ -142,9 +148,9 @@ func (d *recordsBook[K, V]) SetRecord(key K, diff recordDiff[V]) error {
 }
 
 // prune deletes entries from the store that are older than the configured prune expiration.
-// Note that the expiry period is not a strict TTL. Entries that are eligible for deletion may still be present
-// either because the prune function hasn't yet run or because they are still preserved in the in-memory cache after
-// having been deleted from the database.
+// Entries that are eligible for deletion may still be present either because the prune function hasn't yet run or
+// because they are still preserved in the in-memory cache after having been deleted from the database.
+// Such expired entries are filtered out in getRecord
 func (d *recordsBook[K, V]) prune() error {
 	results, err := d.store.Query(d.ctx, query.Query{
 		Prefix: d.dsBaseKey.String(),
@@ -168,7 +174,7 @@ func (d *recordsBook[K, V]) prune() error {
 		if err := v.UnmarshalBinary(result.Value); err != nil {
 			return err
 		}
-		if v.LastUpdated().Add(d.recordExpiry).Before(d.clock.Now()) {
+		if d.hasExpired(v) {
 			if pending > maxPruneBatchSize {
 				if err := batch.Commit(d.ctx); err != nil {
 					return err
@@ -191,6 +197,10 @@ func (d *recordsBook[K, V]) prune() error {
 	return nil
 }
 
+func (d *recordsBook[K, V]) hasExpired(v V) bool {
+	return v.LastUpdated().Add(d.recordExpiry).Before(d.clock.Now())
+}
+
 func (d *recordsBook[K, V]) Close() {
 	d.cancelFn()
 	d.bgTasks.Wait()

op-node/p2p/store/scorebook.go

@@ -12,8 +12,7 @@ import (
 )
 
 const (
-	scoreCacheSize          = 100
-	scoreRecordExpiryPeriod = 24 * time.Hour
+	scoreCacheSize = 100
 )
 
 var scoresBase = ds.NewKey("/peers/scores")
@@ -56,8 +55,8 @@ func peerIDKey(id peer.ID) ds.Key {
 	return ds.NewKey(base32.RawStdEncoding.EncodeToString([]byte(id)))
 }
 
-func newScoreBook(ctx context.Context, logger log.Logger, clock clock.Clock, store ds.Batching) (*scoreBook, error) {
-	book, err := newRecordsBook[peer.ID, *scoreRecord](ctx, logger, clock, store, scoreCacheSize, scoreRecordExpiryPeriod, scoresBase, newScoreRecord, peerIDKey)
+func newScoreBook(ctx context.Context, logger log.Logger, clock clock.Clock, store ds.Batching, retain time.Duration) (*scoreBook, error) {
+	book, err := newRecordsBook[peer.ID, *scoreRecord](ctx, logger, clock, store, scoreCacheSize, retain, scoresBase, newScoreRecord, peerIDKey)
 	if err != nil {
 		return nil, err
 	}

op-node/p2p/store/scorebook_test.go

@@ -81,7 +81,7 @@ func TestPrune(t *testing.T) {
 	logger := testlog.Logger(t, log.LvlInfo)
 	store := sync.MutexWrap(ds.NewMapDatastore())
 	clock := clock.NewDeterministicClock(time.UnixMilli(1000))
-	book, err := newScoreBook(ctx, logger, clock, store)
+	book, err := newScoreBook(ctx, logger, clock, store, 24*time.Hour)
 	require.NoError(t, err)
 
 	hasScoreRecorded := func(id peer.ID) bool {
@@ -135,7 +135,7 @@ func TestPruneMultipleBatches(t *testing.T) {
 	defer cancelFunc()
 	logger := testlog.Logger(t, log.LvlInfo)
 	clock := clock.NewDeterministicClock(time.UnixMilli(1000))
-	book, err := newScoreBook(ctx, logger, clock, sync.MutexWrap(ds.NewMapDatastore()))
+	book, err := newScoreBook(ctx, logger, clock, sync.MutexWrap(ds.NewMapDatastore()), 24*time.Hour)
 	require.NoError(t, err)
 
 	hasScoreRecorded := func(id peer.ID) bool {
@@ -159,6 +159,31 @@ func TestPruneMultipleBatches(t *testing.T) {
 	}
 }
 
+// Check that scores that are eligible for pruning are not returned, even if they haven't yet been removed
+func TestIgnoreOutdatedScores(t *testing.T) {
+	ctx, cancelFunc := context.WithCancel(context.Background())
+	defer cancelFunc()
+	logger := testlog.Logger(t, log.LvlInfo)
+	clock := clock.NewDeterministicClock(time.UnixMilli(1000))
+	retentionPeriod := 24 * time.Hour
+	book, err := newScoreBook(ctx, logger, clock, sync.MutexWrap(ds.NewMapDatastore()), retentionPeriod)
+	require.NoError(t, err)
+
+	require.NoError(t, book.SetScore("a", &GossipScores{Total: 123.45}))
+	clock.AdvanceTime(retentionPeriod + 1)
+
+	// Not available from cache
+	scores, err := book.GetPeerScores("a")
+	require.NoError(t, err)
+	require.Equal(t, scores, PeerScores{})
+
+	book.book.cache.Purge()
+	// Not available from disk
+	scores, err = book.GetPeerScores("a")
+	require.NoError(t, err)
+	require.Equal(t, scores, PeerScores{})
+}
+
 func assertPeerScores(t *testing.T, store ExtendedPeerstore, id peer.ID, expected PeerScores) {
 	result, err := store.GetPeerScores(id)
 	require.NoError(t, err)
@@ -174,8 +199,8 @@ func createPeerstoreWithBacking(t *testing.T, store *sync.MutexDatastore) Extend
 	ps, err := pstoreds.NewPeerstore(context.Background(), store, pstoreds.DefaultOpts())
 	require.NoError(t, err, "Failed to create peerstore")
 	logger := testlog.Logger(t, log.LvlInfo)
-	clock := clock.NewDeterministicClock(time.UnixMilli(100))
-	eps, err := NewExtendedPeerstore(context.Background(), logger, clock, ps, store)
+	c := clock.NewDeterministicClock(time.UnixMilli(100))
+	eps, err := NewExtendedPeerstore(context.Background(), logger, c, ps, store, 24*time.Hour)
 	require.NoError(t, err)
 	t.Cleanup(func() {
 		_ = eps.Close()

proxyd/Dockerfile

-FROM golang:1.18.0-alpine3.15 as builder
+FROM golang:1.20.4-alpine3.18 as builder
 
 ARG GITCOMMIT=docker
 ARG GITDATE=docker
@@ -12,7 +12,7 @@ WORKDIR /app
 
 RUN make proxyd
 
-FROM alpine:3.15
+FROM alpine:3.18
 
 COPY ./proxyd/entrypoint.sh /bin/entrypoint.sh
 

proxyd/backend.go

@@ -374,7 +374,6 @@ func (b *Backend) ForwardRPC(ctx context.Context, res *RPCRes, id string, method
 func (b *Backend) doForward(ctx context.Context, rpcReqs []*RPCReq, isBatch bool) ([]*RPCRes, error) {
 	// we are concerned about network error rates, so we record 1 request independently of how many are in the batch
 	b.networkRequestsSlidingWindow.Incr()
-	RecordBackendNetworkRequestCountSlidingWindow(b, b.networkRequestsSlidingWindow.Count())
 
 	isSingleElementBatch := len(rpcReqs) == 1
 
@@ -391,7 +390,7 @@ func (b *Backend) doForward(ctx context.Context, rpcReqs []*RPCReq, isBatch bool
 	httpReq, err := http.NewRequestWithContext(ctx, "POST", b.rpcURL, bytes.NewReader(body))
 	if err != nil {
 		b.networkErrorsSlidingWindow.Incr()
-		RecordBackendNetworkErrorCountSlidingWindow(b, b.networkErrorsSlidingWindow.Count())
+		RecordBackendNetworkErrorRateSlidingWindow(b, b.ErrorRate())
 		return nil, wrapErr(err, "error creating backend request")
 	}
 
@@ -413,7 +412,7 @@ func (b *Backend) doForward(ctx context.Context, rpcReqs []*RPCReq, isBatch bool
 	httpRes, err := b.client.DoLimited(httpReq)
 	if err != nil {
 		b.networkErrorsSlidingWindow.Incr()
-		RecordBackendNetworkErrorCountSlidingWindow(b, b.networkErrorsSlidingWindow.Count())
+		RecordBackendNetworkErrorRateSlidingWindow(b, b.ErrorRate())
 		return nil, wrapErr(err, "error in backend request")
 	}
 
@@ -432,7 +431,7 @@ func (b *Backend) doForward(ctx context.Context, rpcReqs []*RPCReq, isBatch bool
 	// Alchemy returns a 400 on bad JSONs, so handle that case
 	if httpRes.StatusCode != 200 && httpRes.StatusCode != 400 {
 		b.networkErrorsSlidingWindow.Incr()
-		RecordBackendNetworkErrorCountSlidingWindow(b, b.networkErrorsSlidingWindow.Count())
+		RecordBackendNetworkErrorRateSlidingWindow(b, b.ErrorRate())
 		return nil, fmt.Errorf("response code %d", httpRes.StatusCode)
 	}
 
@@ -440,7 +439,7 @@ func (b *Backend) doForward(ctx context.Context, rpcReqs []*RPCReq, isBatch bool
 	resB, err := io.ReadAll(io.LimitReader(httpRes.Body, b.maxResponseSize))
 	if err != nil {
 		b.networkErrorsSlidingWindow.Incr()
-		RecordBackendNetworkErrorCountSlidingWindow(b, b.networkErrorsSlidingWindow.Count())
+		RecordBackendNetworkErrorRateSlidingWindow(b, b.ErrorRate())
 		return nil, wrapErr(err, "error reading response body")
 	}
 
@@ -458,18 +457,18 @@ func (b *Backend) doForward(ctx context.Context, rpcReqs []*RPCReq, isBatch bool
 			// Infura may return a single JSON-RPC response if, for example, the batch contains a request for an unsupported method
 			if responseIsNotBatched(resB) {
 				b.networkErrorsSlidingWindow.Incr()
-				RecordBackendNetworkErrorCountSlidingWindow(b, b.networkErrorsSlidingWindow.Count())
+				RecordBackendNetworkErrorRateSlidingWindow(b, b.ErrorRate())
 				return nil, ErrBackendUnexpectedJSONRPC
 			}
 			b.networkErrorsSlidingWindow.Incr()
-			RecordBackendNetworkErrorCountSlidingWindow(b, b.networkErrorsSlidingWindow.Count())
+			RecordBackendNetworkErrorRateSlidingWindow(b, b.ErrorRate())
 			return nil, ErrBackendBadResponse
 		}
 	}
 
 	if len(rpcReqs) != len(res) {
 		b.networkErrorsSlidingWindow.Incr()
-		RecordBackendNetworkErrorCountSlidingWindow(b, b.networkErrorsSlidingWindow.Count())
+		RecordBackendNetworkErrorRateSlidingWindow(b, b.ErrorRate())
 		return nil, ErrBackendUnexpectedJSONRPC
 	}
 
@@ -483,6 +482,7 @@ func (b *Backend) doForward(ctx context.Context, rpcReqs []*RPCReq, isBatch bool
 	duration := time.Since(start)
 	b.latencySlidingWindow.Add(float64(duration))
 	RecordBackendNetworkLatencyAverageSlidingWindow(b, time.Duration(b.latencySlidingWindow.Avg()))
+	RecordBackendNetworkErrorRateSlidingWindow(b, b.ErrorRate())
 
 	sortBatchRPCResponse(rpcReqs, res)
 	return res, nil
@@ -490,11 +490,7 @@ func (b *Backend) doForward(ctx context.Context, rpcReqs []*RPCReq, isBatch bool
 
 // IsHealthy checks if the backend is able to serve traffic, based on dynamic parameters
 func (b *Backend) IsHealthy() bool {
-	errorRate := float64(0)
-	// avoid division-by-zero when the window is empty
-	if b.networkRequestsSlidingWindow.Sum() >= 10 {
-		errorRate = b.networkErrorsSlidingWindow.Sum() / b.networkRequestsSlidingWindow.Sum()
-	}
+	errorRate := b.ErrorRate()
 	avgLatency := time.Duration(b.latencySlidingWindow.Avg())
 	if errorRate >= b.maxErrorRateThreshold {
 		return false
@@ -505,6 +501,16 @@ func (b *Backend) IsHealthy() bool {
 	return true
 }
 
+// ErrorRate returns the instant error rate of the backend
+func (b *Backend) ErrorRate() (errorRate float64) {
+	// we only really start counting the error rate after a minimum of 10 requests
+	// this is to avoid false positives when the backend is just starting up
+	if b.networkRequestsSlidingWindow.Sum() >= 10 {
+		errorRate = b.networkErrorsSlidingWindow.Sum() / b.networkRequestsSlidingWindow.Sum()
+	}
+	return errorRate
+}
+
 // IsDegraded checks if the backend is serving traffic in a degraded state (i.e. used as a last resource)
 func (b *Backend) IsDegraded() bool {
 	avgLatency := time.Duration(b.latencySlidingWindow.Avg())
@@ -556,7 +562,11 @@ func (bg *BackendGroup) Forward(ctx context.Context, rpcReqs []*RPCReq, isBatch
 		backends = bg.loadBalancedConsensusGroup()
 
 		// We also rewrite block tags to enforce compliance with consensus
-		rctx := RewriteContext{latest: bg.Consensus.GetConsensusBlockNumber()}
+		rctx := RewriteContext{
+			latest:    bg.Consensus.GetLatestBlockNumber(),
+			finalized: bg.Consensus.GetFinalizedBlockNumber(),
+			safe:      bg.Consensus.GetSafeBlockNumber(),
+		}
 
 		for i, req := range rpcReqs {
 			res := RPCRes{JSONRPC: JSONRPCVersion, ID: req.ID}

proxyd/consensus_tracker.go

@@ -13,35 +13,68 @@ import (
 // ConsensusTracker abstracts how we store and retrieve the current consensus
 // allowing it to be stored locally in-memory or in a shared Redis cluster
 type ConsensusTracker interface {
-	GetConsensusBlockNumber() hexutil.Uint64
-	SetConsensusBlockNumber(blockNumber hexutil.Uint64)
+	GetLatestBlockNumber() hexutil.Uint64
+	SetLatestBlockNumber(blockNumber hexutil.Uint64)
+	GetFinalizedBlockNumber() hexutil.Uint64
+	SetFinalizedBlockNumber(blockNumber hexutil.Uint64)
+	GetSafeBlockNumber() hexutil.Uint64
+	SetSafeBlockNumber(blockNumber hexutil.Uint64)
 }
 
 // InMemoryConsensusTracker store and retrieve in memory, async-safe
 type InMemoryConsensusTracker struct {
-	consensusBlockNumber hexutil.Uint64
+	latestBlockNumber    hexutil.Uint64
+	finalizedBlockNumber hexutil.Uint64
+	safeBlockNumber      hexutil.Uint64
 	mutex                sync.Mutex
 }
 
 func NewInMemoryConsensusTracker() ConsensusTracker {
 	return &InMemoryConsensusTracker{
-		consensusBlockNumber: 0,
-		mutex:                sync.Mutex{},
+		mutex: sync.Mutex{},
 	}
 }
 
-func (ct *InMemoryConsensusTracker) GetConsensusBlockNumber() hexutil.Uint64 {
+func (ct *InMemoryConsensusTracker) GetLatestBlockNumber() hexutil.Uint64 {
 	defer ct.mutex.Unlock()
 	ct.mutex.Lock()
 
-	return ct.consensusBlockNumber
+	return ct.latestBlockNumber
 }
 
-func (ct *InMemoryConsensusTracker) SetConsensusBlockNumber(blockNumber hexutil.Uint64) {
+func (ct *InMemoryConsensusTracker) SetLatestBlockNumber(blockNumber hexutil.Uint64) {
 	defer ct.mutex.Unlock()
 	ct.mutex.Lock()
 
-	ct.consensusBlockNumber = blockNumber
+	ct.latestBlockNumber = blockNumber
+}
+
+func (ct *InMemoryConsensusTracker) GetFinalizedBlockNumber() hexutil.Uint64 {
+	defer ct.mutex.Unlock()
+	ct.mutex.Lock()
+
+	return ct.finalizedBlockNumber
+}
+
+func (ct *InMemoryConsensusTracker) SetFinalizedBlockNumber(blockNumber hexutil.Uint64) {
+	defer ct.mutex.Unlock()
+	ct.mutex.Lock()
+
+	ct.finalizedBlockNumber = blockNumber
+}
+
+func (ct *InMemoryConsensusTracker) GetSafeBlockNumber() hexutil.Uint64 {
+	defer ct.mutex.Unlock()
+	ct.mutex.Lock()
+
+	return ct.safeBlockNumber
+}
+
+func (ct *InMemoryConsensusTracker) SetSafeBlockNumber(blockNumber hexutil.Uint64) {
+	defer ct.mutex.Unlock()
+	ct.mutex.Lock()
+
+	ct.safeBlockNumber = blockNumber
 }
 
 // RedisConsensusTracker uses a Redis `client` to store and retrieve consensus, async-safe
@@ -59,14 +92,29 @@ func NewRedisConsensusTracker(ctx context.Context, r *redis.Client, namespace st
 	}
 }
 
-func (ct *RedisConsensusTracker) key() string {
-	return fmt.Sprintf("consensus_latest_block:%s", ct.backendGroup)
+func (ct *RedisConsensusTracker) key(tag string) string {
+	return fmt.Sprintf("consensus:%s:%s", ct.backendGroup, tag)
 }
 
-func (ct *RedisConsensusTracker) GetConsensusBlockNumber() hexutil.Uint64 {
-	return hexutil.Uint64(hexutil.MustDecodeUint64(ct.client.Get(ct.ctx, ct.key()).Val()))
+func (ct *RedisConsensusTracker) GetLatestBlockNumber() hexutil.Uint64 {
+	return hexutil.Uint64(hexutil.MustDecodeUint64(ct.client.Get(ct.ctx, ct.key("latest")).Val()))
+}
+
+func (ct *RedisConsensusTracker) SetLatestBlockNumber(blockNumber hexutil.Uint64) {
+	ct.client.Set(ct.ctx, ct.key("latest"), blockNumber, 0)
+}
+
+func (ct *RedisConsensusTracker) GetFinalizedBlockNumber() hexutil.Uint64 {
+	return hexutil.Uint64(hexutil.MustDecodeUint64(ct.client.Get(ct.ctx, ct.key("finalized")).Val()))
+}
+
+func (ct *RedisConsensusTracker) SetFinalizedBlockNumber(blockNumber hexutil.Uint64) {
+	ct.client.Set(ct.ctx, ct.key("finalized"), blockNumber, 0)
+}
+func (ct *RedisConsensusTracker) GetSafeBlockNumber() hexutil.Uint64 {
+	return hexutil.Uint64(hexutil.MustDecodeUint64(ct.client.Get(ct.ctx, ct.key("safe")).Val()))
 }
 
-func (ct *RedisConsensusTracker) SetConsensusBlockNumber(blockNumber hexutil.Uint64) {
-	ct.client.Set(ct.ctx, ct.key(), blockNumber, 0)
+func (ct *RedisConsensusTracker) SetSafeBlockNumber(blockNumber hexutil.Uint64) {
+	ct.client.Set(ct.ctx, ct.key("safe"), blockNumber, 0)
 }

proxyd/example.config.toml

@@ -93,8 +93,8 @@ backends = ["infura"]
 # consensus_ban_period = "1m"
 # Maximum delay for update the backend, default 30s
 # consensus_max_update_threshold = "20s"
-# Maximum block lag, default 50
-# consensus_max_block_lag = 10
+# Maximum block lag, default 8
+# consensus_max_block_lag = 16
 # Minimum peer count, default 3
 # consensus_min_peer_count = 4
 

proxyd/integration_tests/testdata/consensus.toml

@@ -18,7 +18,7 @@ consensus_aware = true
 consensus_handler = "noop" # allow more control over the consensus poller for tests
 consensus_ban_period = "1m"
 consensus_max_update_threshold = "2m"
-consensus_max_block_lag = 50
+consensus_max_block_lag = 8
 consensus_min_peer_count = 4
 
 [rpc_method_mappings]

proxyd/integration_tests/testdata/consensus_responses.yml

@@ -26,40 +26,161 @@
       "jsonrpc": "2.0",
       "id": 67,
       "result": {
-        "hash": "hash1",
-        "number": "0x1"
+        "hash": "hash_0x101",
+        "number": "0x101"
       }
     }
 - method: eth_getBlockByNumber
-  block: 0x1
+  block: 0x101
   response: >
     {
       "jsonrpc": "2.0",
       "id": 67,
       "result": {
-        "hash": "hash1",
-        "number": "0x1"
+        "hash": "hash_0x101",
+        "number": "0x101"
       }
     }
 - method: eth_getBlockByNumber
-  block: 0x2
+  block: 0x102
   response: >
     {
       "jsonrpc": "2.0",
       "id": 67,
       "result": {
-        "hash": "hash2",
-        "number": "0x2"
+        "hash": "hash_0x102",
+        "number": "0x102"
       }
     }
 - method: eth_getBlockByNumber
-  block: 0x3
+  block: 0x103
   response: >
     {
       "jsonrpc": "2.0",
       "id": 67,
       "result": {
-        "hash": "hash3",
-        "number": "0x3"
+        "hash": "hash_0x103",
+        "number": "0x103"
+      }
+    }
+- method: eth_getBlockByNumber
+  block: 0x10a
+  response: >
+    {
+      "jsonrpc": "2.0",
+      "id": 67,
+      "result": {
+        "hash": "hash_0x10a",
+        "number": "0x10a"
+      }
+    }
+- method: eth_getBlockByNumber
+  block: 0x132
+  response: >
+    {
+      "jsonrpc": "2.0",
+      "id": 67,
+      "result": {
+        "hash": "hash_0x132",
+        "number": "0x132"
+      }
+    }
+- method: eth_getBlockByNumber
+  block: 0x133
+  response: >
+    {
+      "jsonrpc": "2.0",
+      "id": 67,
+      "result": {
+        "hash": "hash_0x133",
+        "number": "0x133"
+      }
+    }
+- method: eth_getBlockByNumber
+  block: 0x134
+  response: >
+    {
+      "jsonrpc": "2.0",
+      "id": 67,
+      "result": {
+        "hash": "hash_0x134",
+        "number": "0x134"
+      }
+    }
+- method: eth_getBlockByNumber
+  block: 0x200
+  response: >
+    {
+      "jsonrpc": "2.0",
+      "id": 67,
+      "result": {
+        "hash": "hash_0x200",
+        "number": "0x200"
+      }
+    }
+- method: eth_getBlockByNumber
+  block: 0x91
+  response: >
+    {
+      "jsonrpc": "2.0",
+      "id": 67,
+      "result": {
+        "hash": "hash_0x91",
+        "number": "0x91"
+      }
+    }
+- method: eth_getBlockByNumber
+  block: safe
+  response: >
+    {
+      "jsonrpc": "2.0",
+      "id": 67,
+      "result": {
+        "hash": "hash_0xe1",
+        "number": "0xe1"
+      }
+    }
+- method: eth_getBlockByNumber
+  block: 0xe1
+  response: >
+    {
+      "jsonrpc": "2.0",
+      "id": 67,
+      "result": {
+        "hash": "hash_0xe1",
+        "number": "0xe1"
+      }
+    }
+- method: eth_getBlockByNumber
+  block: finalized
+  response: >
+    {
+      "jsonrpc": "2.0",
+      "id": 67,
+      "result": {
+        "hash": "hash_0xc1",
+        "number": "0xc1"
+      }
+    }
+- method: eth_getBlockByNumber
+  block: 0xc1
+  response: >
+    {
+      "jsonrpc": "2.0",
+      "id": 67,
+      "result": {
+        "hash": "hash_0xc1",
+        "number": "0xc1"
+      }
+    }
+- method: eth_getBlockByNumber
+  block: 0xd1
+  response: >
+    {
+      "jsonrpc": "2.0",
+      "id": 67,
+      "result": {
+        "hash": "hash_0xd1",
+        "number": "0xd1"
       }
     }

proxyd/metrics.go

@@ -246,6 +246,22 @@ var (
 		"backend_group_name",
 	})
 
+	consensusSafeBlock = promauto.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: MetricsNamespace,
+		Name:      "group_consensus_safe_block",
+		Help:      "Consensus safe block",
+	}, []string{
+		"backend_group_name",
+	})
+
+	consensusFinalizedBlock = promauto.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: MetricsNamespace,
+		Name:      "group_consensus_finalized_block",
+		Help:      "Consensus finalized block",
+	}, []string{
+		"backend_group_name",
+	})
+
 	backendLatestBlockBackend = promauto.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: MetricsNamespace,
 		Name:      "backend_latest_block",
@@ -254,6 +270,30 @@ var (
 		"backend_name",
 	})
 
+	backendSafeBlockBackend = promauto.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: MetricsNamespace,
+		Name:      "backend_safe_block",
+		Help:      "Current safe block observed per backend",
+	}, []string{
+		"backend_name",
+	})
+
+	backendFinalizedBlockBackend = promauto.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: MetricsNamespace,
+		Name:      "backend_finalized_block",
+		Help:      "Current finalized block observed per backend",
+	}, []string{
+		"backend_name",
+	})
+
+	backendUnexpectedBlockTagsBackend = promauto.NewGaugeVec(prometheus.GaugeOpts{
+		Namespace: MetricsNamespace,
+		Name:      "backend_unexpected_block_tags",
+		Help:      "Bool gauge for unexpected block tags",
+	}, []string{
+		"backend_name",
+	})
+
 	consensusGroupCount = promauto.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: MetricsNamespace,
 		Name:      "group_consensus_count",
@@ -318,18 +358,10 @@ var (
 		"backend_name",
 	})
 
-	networkErrorCountBackend = promauto.NewGaugeVec(prometheus.GaugeOpts{
+	networkErrorRateBackend = promauto.NewGaugeVec(prometheus.GaugeOpts{
 		Namespace: MetricsNamespace,
-		Name:      "backend_net_error_count",
-		Help:      "Network error count per backend",
-	}, []string{
-		"backend_name",
-	})
-
-	requestCountBackend = promauto.NewGaugeVec(prometheus.GaugeOpts{
-		Namespace: MetricsNamespace,
-		Name:      "backend_request_count",
-		Help:      "Request count per backend",
+		Name:      "backend_error_rate",
+		Help:      "Request error rate per backend",
 	}, []string{
 		"backend_name",
 	})
@@ -402,6 +434,14 @@ func RecordGroupConsensusLatestBlock(group *BackendGroup, blockNumber hexutil.Ui
 	consensusLatestBlock.WithLabelValues(group.Name).Set(float64(blockNumber))
 }
 
+func RecordGroupConsensusSafeBlock(group *BackendGroup, blockNumber hexutil.Uint64) {
+	consensusSafeBlock.WithLabelValues(group.Name).Set(float64(blockNumber))
+}
+
+func RecordGroupConsensusFinalizedBlock(group *BackendGroup, blockNumber hexutil.Uint64) {
+	consensusFinalizedBlock.WithLabelValues(group.Name).Set(float64(blockNumber))
+}
+
 func RecordGroupConsensusCount(group *BackendGroup, count int) {
 	consensusGroupCount.WithLabelValues(group.Name).Set(float64(count))
 }
@@ -418,12 +458,20 @@ func RecordBackendLatestBlock(b *Backend, blockNumber hexutil.Uint64) {
 	backendLatestBlockBackend.WithLabelValues(b.Name).Set(float64(blockNumber))
 }
 
+func RecordBackendSafeBlock(b *Backend, blockNumber hexutil.Uint64) {
+	backendSafeBlockBackend.WithLabelValues(b.Name).Set(float64(blockNumber))
+}
+
+func RecordBackendFinalizedBlock(b *Backend, blockNumber hexutil.Uint64) {
+	backendFinalizedBlockBackend.WithLabelValues(b.Name).Set(float64(blockNumber))
+}
+
+func RecordBackendUnexpectedBlockTags(b *Backend, unexpected bool) {
+	backendUnexpectedBlockTagsBackend.WithLabelValues(b.Name).Set(boolToFloat64(unexpected))
+}
+
 func RecordConsensusBackendBanned(b *Backend, banned bool) {
-	v := float64(0)
-	if banned {
-		v = float64(1)
-	}
-	consensusBannedBackends.WithLabelValues(b.Name).Set(v)
+	consensusBannedBackends.WithLabelValues(b.Name).Set(boolToFloat64(banned))
 }
 
 func RecordConsensusBackendPeerCount(b *Backend, peerCount uint64) {
@@ -431,11 +479,7 @@ func RecordConsensusBackendPeerCount(b *Backend, peerCount uint64) {
 }
 
 func RecordConsensusBackendInSync(b *Backend, inSync bool) {
-	v := float64(0)
-	if inSync {
-		v = float64(1)
-	}
-	consensusInSyncBackend.WithLabelValues(b.Name).Set(v)
+	consensusInSyncBackend.WithLabelValues(b.Name).Set(boolToFloat64(inSync))
 }
 
 func RecordConsensusBackendUpdateDelay(b *Backend, delay time.Duration) {
@@ -446,10 +490,13 @@ func RecordBackendNetworkLatencyAverageSlidingWindow(b *Backend, avgLatency time
 	avgLatencyBackend.WithLabelValues(b.Name).Set(float64(avgLatency.Milliseconds()))
 }
 
-func RecordBackendNetworkRequestCountSlidingWindow(b *Backend, count uint) {
-	requestCountBackend.WithLabelValues(b.Name).Set(float64(count))
+func RecordBackendNetworkErrorRateSlidingWindow(b *Backend, rate float64) {
+	networkErrorRateBackend.WithLabelValues(b.Name).Set(rate)
 }
 
-func RecordBackendNetworkErrorCountSlidingWindow(b *Backend, count uint) {
-	networkErrorCountBackend.WithLabelValues(b.Name).Set(float64(count))
+func boolToFloat64(b bool) float64 {
+	if b {
+		return 1
+	}
+	return 0
 }

proxyd/rewriter.go

@@ -9,7 +9,9 @@ import (
 )
 
 type RewriteContext struct {
-	latest hexutil.Uint64
+	latest    hexutil.Uint64
+	safe      hexutil.Uint64
+	finalized hexutil.Uint64
 }
 
 type RewriteResult uint8
@@ -180,11 +182,13 @@ func rewriteTag(rctx RewriteContext, current string) (string, bool, error) {
 	}
 
 	switch *bnh.BlockNumber {
-	case rpc.SafeBlockNumber,
-		rpc.FinalizedBlockNumber,
-		rpc.PendingBlockNumber,
+	case rpc.PendingBlockNumber,
 		rpc.EarliestBlockNumber:
 		return current, false, nil
+	case rpc.FinalizedBlockNumber:
+		return rctx.finalized.String(), true, nil
+	case rpc.SafeBlockNumber:
+		return rctx.safe.String(), true, nil
 	case rpc.LatestBlockNumber:
 		return rctx.latest.String(), true, nil
 	default:

proxyd/rewriter_test.go

@@ -326,33 +326,33 @@ func TestRewriteRequest(t *testing.T) {
 		{
 			name: "eth_getBlockByNumber finalized",
 			args: args{
-				rctx: RewriteContext{latest: hexutil.Uint64(100)},
+				rctx: RewriteContext{latest: hexutil.Uint64(100), finalized: hexutil.Uint64(55)},
 				req:  &RPCReq{Method: "eth_getBlockByNumber", Params: mustMarshalJSON([]string{"finalized"})},
 				res:  nil,
 			},
-			expected: RewriteNone,
+			expected: RewriteOverrideRequest,
 			check: func(t *testing.T, args args) {
 				var p []string
 				err := json.Unmarshal(args.req.Params, &p)
 				require.Nil(t, err)
 				require.Equal(t, 1, len(p))
-				require.Equal(t, "finalized", p[0])
+				require.Equal(t, hexutil.Uint64(55).String(), p[0])
 			},
 		},
 		{
 			name: "eth_getBlockByNumber safe",
 			args: args{
-				rctx: RewriteContext{latest: hexutil.Uint64(100)},
+				rctx: RewriteContext{latest: hexutil.Uint64(100), safe: hexutil.Uint64(50)},
 				req:  &RPCReq{Method: "eth_getBlockByNumber", Params: mustMarshalJSON([]string{"safe"})},
 				res:  nil,
 			},
-			expected: RewriteNone,
+			expected: RewriteOverrideRequest,
 			check: func(t *testing.T, args args) {
 				var p []string
 				err := json.Unmarshal(args.req.Params, &p)
 				require.Nil(t, err)
 				require.Equal(t, 1, len(p))
-				require.Equal(t, "safe", p[0])
+				require.Equal(t, hexutil.Uint64(50).String(), p[0])
 			},
 		},
 		{

proxyd/tools/mockserver/handler/handler.go

@@ -95,7 +95,7 @@ func (mh *MockedHandler) Handler(w http.ResponseWriter, req *http.Request) {
 	resBody := ""
 	if batched {
 		resBody = "[" + strings.Join(responses, ",") + "]"
-	} else {
+	} else if len(responses) > 0 {
 		resBody = responses[0]
 	}