fix(rollup-relayer): better handle commit batch revert (#981)

Co-authored-by: colinlyguo <[email protected]>
Co-authored-by: HAOYUatHZ <[email protected]>

common/version/version.go

@@ -5,7 +5,7 @@ import (
 	"runtime/debug"
 )
 
-var tag = "v4.3.30"
+var tag = "v4.3.31"
 
 var commit = func() string {
 	if info, ok := debug.ReadBuildInfo(); ok {

rollup/internal/controller/relayer/l2_relayer.go

@@ -307,12 +307,12 @@ func (r *Layer2Relayer) ProcessGasPriceOracle() {
 // ProcessPendingBatches processes the pending batches by sending commitBatch transactions to layer 1.
 func (r *Layer2Relayer) ProcessPendingBatches() {
 	// get pending batches from database in ascending order by their index.
-	pendingBatches, err := r.batchOrm.GetPendingBatches(r.ctx, 5)
+	batches, err := r.batchOrm.GetFailedAndPendingBatches(r.ctx, 5)
 	if err != nil {
 		log.Error("Failed to fetch pending L2 batches", "err", err)
 		return
 	}
-	for _, batch := range pendingBatches {
+	for _, batch := range batches {
 		r.metrics.rollupL2RelayerProcessPendingBatchTotal.Inc()
 		// get current header and parent header.
 		currentBatchHeader, err := types.DecodeBatchHeader(batch.BatchHeader)
@@ -327,6 +327,12 @@ func (r *Layer2Relayer) ProcessPendingBatches() {
 				log.Error("Failed to get parent batch header", "index", batch.Index-1, "error", err)
 				return
 			}
+
+			if types.RollupStatus(parentBatch.RollupStatus) == types.RollupCommitFailed {
+				log.Error("Previous batch commit failed, halting further committing",
+					"index", parentBatch.Index, "tx hash", parentBatch.CommitTxHash)
+				return
+			}
 		}
 
 		// get the chunks for the batch
@@ -371,6 +377,11 @@ func (r *Layer2Relayer) ProcessPendingBatches() {
 		// send transaction
 		txID := batch.Hash + "-commit"
 		fallbackGasLimit := uint64(float64(batch.TotalL1CommitGas) * r.cfg.L1CommitGasLimitMultiplier)
+		if types.RollupStatus(batch.RollupStatus) == types.RollupCommitFailed {
+			// use eth_estimateGas if this batch has been committed failed.
+			fallbackGasLimit = 0
+			log.Warn("Batch commit previously failed, using eth_estimateGas for the re-submission", "hash", batch.Hash)
+		}
 		txHash, err := r.commitSender.SendTransaction(txID, &r.cfg.RollupContractAddress, big.NewInt(0), calldata, fallbackGasLimit)
 		if err != nil {
 			log.Error(

rollup/internal/orm/batch.go

@@ -192,22 +192,22 @@ func (o *Batch) GetRollupStatusByHashList(ctx context.Context, hashes []string)
 	return statuses, nil
 }
 
-// GetPendingBatches retrieves pending batches up to the specified limit.
+// GetFailedAndPendingBatches retrieves batches with failed or pending status up to the specified limit.
 // The returned batches are sorted in ascending order by their index.
-func (o *Batch) GetPendingBatches(ctx context.Context, limit int) ([]*Batch, error) {
+func (o *Batch) GetFailedAndPendingBatches(ctx context.Context, limit int) ([]*Batch, error) {
 	if limit <= 0 {
 		return nil, errors.New("limit must be greater than zero")
 	}
 
 	db := o.db.WithContext(ctx)
 	db = db.Model(&Batch{})
-	db = db.Where("rollup_status = ?", types.RollupPending)
+	db = db.Where("rollup_status = ? OR rollup_status = ?", types.RollupCommitFailed, types.RollupPending)
 	db = db.Order("index ASC")
 	db = db.Limit(limit)
 
 	var batches []*Batch
 	if err := db.Find(&batches).Error; err != nil {
-		return nil, fmt.Errorf("Batch.GetPendingBatches error: %w", err)
+		return nil, fmt.Errorf("Batch.GetFailedAndPendingBatches error: %w", err)
 	}
 	return batches, nil
 }

rollup/internal/orm/orm_test.go

@@ -258,14 +258,17 @@ func TestBatchOrm(t *testing.T) {
 	assert.NoError(t, err)
 	assert.Equal(t, uint64(2), count)
 
-	pendingBatches, err := batchOrm.GetPendingBatches(context.Background(), 100)
+	err = batchOrm.UpdateRollupStatus(context.Background(), batchHash1, types.RollupCommitFailed)
+	assert.NoError(t, err)
+
+	pendingBatches, err := batchOrm.GetFailedAndPendingBatches(context.Background(), 100)
 	assert.NoError(t, err)
 	assert.Equal(t, 2, len(pendingBatches))
 
 	rollupStatus, err := batchOrm.GetRollupStatusByHashList(context.Background(), []string{batchHash1, batchHash2})
 	assert.NoError(t, err)
 	assert.Equal(t, 2, len(rollupStatus))
-	assert.Equal(t, types.RollupPending, rollupStatus[0])
+	assert.Equal(t, types.RollupCommitFailed, rollupStatus[0])
 	assert.Equal(t, types.RollupPending, rollupStatus[1])
 
 	err = batchOrm.UpdateProvingStatus(context.Background(), batchHash2, types.ProvingTaskVerified)