refactor(pruner): remove persisted height coordination (#494)

Author: freemans13

services/pruner/blob_deletion_test.go

@@ -217,7 +217,7 @@ func TestBlobDeletionSchedulingAndExecution(t *testing.T) {
 	// Set up mock settings
 	server.settings = &settings.Settings{
 		Pruner: settings.PrunerSettings{
-			BlobDeletionEnabled:      true,
+			SkipBlobDeletion:         false,
 			BlobDeletionSafetyWindow: 0,
 			BlobDeletionBatchSize:    100,
 			BlobDeletionMaxRetries:   3,
@@ -377,7 +377,7 @@ func TestBlobDeletionIdempotency(t *testing.T) {
 
 	server.settings = &settings.Settings{
 		Pruner: settings.PrunerSettings{
-			BlobDeletionEnabled:    true,
+			SkipBlobDeletion:       false,
 			BlobDeletionBatchSize:  100,
 			BlobDeletionMaxRetries: 3,
 		},

services/pruner/blob_deletion_worker.go

@@ -39,7 +39,7 @@ func (s *Server) blobDeletionWorker() {
 }
 
 func (s *Server) processBlobDeletionsAtHeight(height uint32, blockHash *chainhash.Hash) {
-	if !s.settings.Pruner.BlobDeletionEnabled {
+	if s.settings.Pruner.SkipBlobDeletion {
 		return
 	}
 

services/pruner/server.go

@@ -127,9 +127,6 @@ func (s *Server) Init(ctx context.Context) error {
 		return errors.NewServiceError("pruner service not available from UTXO store")
 	}
 
-	// Set persisted height getter for block persister coordination
-	s.prunerService.SetPersistedHeightGetter(s.GetLastPersistedHeight)
-
 	// Validate block trigger mode
 	blockTrigger := s.settings.Pruner.BlockTrigger
 	if blockTrigger != settings.PrunerBlockTriggerOnBlockPersisted && blockTrigger != settings.PrunerBlockTriggerOnBlockMined {
@@ -234,31 +231,18 @@ func (s *Server) Init(ctx context.Context) error {
 					continue
 				}
 
-				// Get height from block assembly state
-				state, err := s.blockAssemblyClient.GetBlockAssemblyState(ctx)
+				// Get height from blockchain client using the block hash
+				// This is the authoritative source and avoids race conditions with Block Assembly state updates
+				header, meta, err := s.blockchainClient.GetBlockHeader(ctx, blockHash)
 				if err != nil {
-					s.logger.Debugf("Failed to get block assembly state on Block notification: %v", err)
+					s.logger.Debugf("Failed to get block header for Block notification hash %s: %v", blockHash, err)
 					continue
 				}
-
-				// If block assembly hasn't initialized yet (height=0), get height from blockchain
-				// This handles the edge case where Block notifications arrive before block assembly
-				// has processed its first block
-				var height uint32
-				if state.CurrentHeight == 0 {
-					header, meta, err := s.blockchainClient.GetBlockHeader(ctx, blockHash)
-					if err != nil {
-						s.logger.Debugf("Failed to get block header for Block notification hash %s: %v", blockHash, err)
-						continue
-					}
-					if header == nil || meta == nil {
-						s.logger.Debugf("Block notification for hash %s has no header/meta", blockHash)
-						continue
-					}
-					height = meta.Height
-				} else {
-					height = state.CurrentHeight
+				if header == nil || meta == nil {
+					s.logger.Debugf("Block notification for hash %s has no header/meta", blockHash)
+					continue
 				}
+				height := meta.Height
 
 				// Queue pruning request immediately - processor will wait for mined_set if block assembly is running
 				if height > s.lastProcessedHeight.Load() {
@@ -426,13 +410,6 @@ func (s *Server) HealthGRPC(ctx context.Context, _ *pruner_api.EmptyMessage) (*p
 	}, errors.WrapGRPC(err)
 }
 
-// GetLastPersistedHeight returns the last known block height that has been persisted
-// by the block persister service. This is used to coordinate cleanup operations to
-// avoid deleting data that the block persister still needs.
-func (s *Server) GetLastPersistedHeight() uint32 {
-	return s.lastPersistedHeight.Load()
-}
-
 // SetBlobDeletionObserver sets an optional observer for blob deletion completion events.
 // This is primarily used for testing to synchronize test execution with deletion processing.
 func (s *Server) SetBlobDeletionObserver(observer BlobDeletionObserver) {

services/pruner/worker.go

@@ -240,7 +240,7 @@ func (s *Server) prunerProcessor(ctx context.Context) {
 
 			// Phase 1: Preserve parents of old unmined transactions
 			// This must run before Phase 2 to protect parents from deletion
-			if s.utxoStore != nil {
+			if s.utxoStore != nil && !s.settings.Pruner.SkipPreserveParents {
 				hashStr := "<unknown>"
 				if latestReq.BlockHash != nil {
 					hashStr = latestReq.BlockHash.String()
@@ -260,6 +260,12 @@ func (s *Server) prunerProcessor(ctx context.Context) {
 						prunerUpdatingParents.Add(float64(count))
 					}
 				}
+			} else if s.settings.Pruner.SkipPreserveParents {
+				hashStr := "<unknown>"
+				if latestReq.BlockHash != nil {
+					hashStr = latestReq.BlockHash.String()
+				}
+				s.logger.Infof("[pruner][%s:%d] phase 1: skipped (pruner_skipPreserveParents=true)", hashStr, latestReq.Height)
 			}
 
 			// Phase 2: DAH pruning (deletion)

settings/pruner_settings.go

@@ -18,10 +18,11 @@ type PrunerSettings struct {
 	UTXOProgressLogInterval         time.Duration `key:"pruner_utxoProgressLogInterval" desc:"Interval for logging pruning progress" default:"30s" category:"Pruner" usage:"How often to log progress during UTXO pruning" type:"duration" longdesc:"### Purpose\nSets how often to log progress during UTXO pruning operations.\n\n### How It Works\nDuring long-running pruning operations, progress is logged at this interval to help operators monitor the operation without flooding logs.\n\n### Recommendations\n- **30s** (default) - Regular updates without excessive logging\n- Decrease for more frequent updates during debugging\n- Increase for quieter logs in stable environments"`
 	UTXOPartitionQueries            int           `key:"pruner_utxoPartitionQueries" desc:"Number of parallel Aerospike partition queries" default:"0" category:"Pruner" usage:"0 = auto-detect based on CPU cores" type:"int" longdesc:"### Purpose\nControls the number of parallel partition workers when scanning Aerospike for prunable records.\n\n### How It Works\n- Aerospike keyspace is divided into 4096 partitions\n- This setting controls how many workers scan partitions in parallel\n- Each worker processes a range of partitions independently\n- Partition-based scanning achieves up to 100x performance improvement over sequential queries\n\n### Values\n- **0** (default) - Auto-detect based on CPU cores and Aerospike query-threads-limit\n- **N > 0** - Fixed number of partition workers (capped at 4096)\n\n### Trade-offs\n| Setting | Benefit | Drawback |\n|---------|---------|----------|\n| Higher | Faster scanning | More Aerospike load, more connections |\n| Lower | Reduced cluster pressure | Slower pruning |\n\n### Recommendations\n- Use **0** for automatic scaling based on available resources\n- Set explicitly to match your Aerospike cluster's capacity"`
 	UTXOSetTTL                      bool          `key:"pruner_utxoSetTTL" desc:"Use TTL expiration instead of hard delete" default:"false" category:"Pruner" usage:"Set record TTL instead of deleting" type:"bool" longdesc:"### Purpose\nControls whether the pruner uses TTL-based expiration instead of hard deletes for UTXO records.\n\n### How It Works\n- When enabled, instead of deleting records directly, the pruner sets the Aerospike record TTL to 1 second\n- Aerospike's nsup (namespace supervisor) thread handles the actual expiration in the background\n- This produces optimized tombstones and reduces write amplification compared to hard deletes\n- The record becomes inaccessible within 1 second and is fully cleaned up by nsup\n\n### Trade-offs\n| Setting | Benefit | Drawback |\n|---------|---------|----------|\n| true | Reduced write amplification, optimized tombstones | Records persist ~1 second longer |\n| false (default) | Immediate deletion | Higher write amplification, standard tombstones |\n\n### Recommendations\n- **false** (default) - Standard hard delete behavior\n- **true** - Recommended for high-throughput environments where nsup-managed expiration is preferred"`
-	BlobDeletionEnabled             bool          `key:"pruner_blobDeletionEnabled" desc:"Enable blob deletion scheduling" default:"true" category:"Pruner" usage:"Enable deletion of expired blobs" type:"bool" longdesc:"### Purpose\nEnables scheduled deletion of blob store data (transactions and subtrees) based on Delete-At-Height (DAH) values.\n\n### How It Works\n- When enabled, pruner schedules deletions for blobs that have reached their DAH height\n- Works with all blob store types (file, S3, memory, etc.)\n- Deletion is triggered after BlobDeletionSafetyWindow blocks past the persister height\n\n### Recommendations\n- **true** (default) - Enable for normal operations\n- **false** - Disable if you want to retain all blob data indefinitely"`
+	SkipBlobDeletion                bool          `key:"pruner_skipBlobDeletion" desc:"Skip blob deletion scheduling" default:"false" category:"Pruner" usage:"Skip deletion of expired blobs" type:"bool" longdesc:"### Purpose\nSkips scheduled deletion of blob store data (transactions and subtrees) based on Delete-At-Height (DAH) values.\n\n### How It Works\n- When enabled, pruner skips deletions for blobs that have reached their DAH height\n- Works with all blob store types (file, S3, memory, etc.)\n- Deletion would normally be triggered after BlobDeletionSafetyWindow blocks past the persister height\n\n### Recommendations\n- **false** (default) - Enable blob deletion for normal operations\n- **true** - Skip blob deletion if you want to retain all blob data indefinitely"`
 	BlobDeletionSafetyWindow        uint32        `key:"pruner_blobDeletionSafetyWindow" desc:"Blocks to wait after persister before deletion" default:"10" category:"Pruner" usage:"Safety margin for blob deletion" type:"uint32" longdesc:"### Purpose\nNumber of blocks to wait after Block Persister height before deleting blobs.\n\n### How It Works\nProvides a safety margin to ensure persisted blocks are stable before deleting their associated blobs. Prevents deletion of data that might be needed during reorg scenarios.\n\n### Recommendations\n- **10** (default) - Good balance between storage and safety\n- Increase for deeper reorg protection\n- Decrease to free storage faster (not recommended)"`
 	BlobDeletionBatchSize           int           `key:"pruner_blobDeletionBatchSize" desc:"Maximum deletions per trigger" default:"1000" category:"Pruner" usage:"Limits deletions per pruning cycle" type:"int" longdesc:"### Purpose\nMaximum number of blob deletions to process per pruning trigger.\n\n### How It Works\nLimits the number of deletions per cycle to prevent overwhelming the blob store and database. Remaining deletions are processed in subsequent triggers.\n\n### Recommendations\n- **1000** (default) - Good throughput without excessive load\n- Increase for faster cleanup if system can handle it\n- Decrease to reduce deletion load"`
 	BlobDeletionMaxRetries          int           `key:"pruner_blobDeletionMaxRetries" desc:"Maximum retry attempts for failed deletions" default:"3" category:"Pruner" usage:"Retries for transient failures" type:"int" longdesc:"### Purpose\nMaximum number of retry attempts for failed blob deletions.\n\n### How It Works\nWhen a blob deletion fails (network error, temporary unavailability), the pruner retries up to this many times before logging an error.\n\n### Recommendations\n- **3** (default) - Good balance for transient failures\n- Increase for unreliable storage backends\n- Decrease for faster failure detection"`
+	SkipPreserveParents             bool          `key:"pruner_skipPreserveParents" desc:"Skip Phase 1: preserve parents of unmined transactions" default:"false" category:"Pruner" usage:"Skip parent preservation phase" type:"bool" longdesc:"### Purpose\nSkips Phase 1 of pruning which preserves parent transactions of old unmined transactions.\n\n### How It Works\nWhen enabled, the pruner skips calling PreserveParentsOfOldUnminedTransactions.\nThis means parent transactions will not be protected from deletion even if they have unmined children.\n\n### Trade-offs\n| Setting | Benefit | Drawback |\n|---------|---------|----------|\n| false (default) | Parents preserved for unmined tx resubmission | Additional processing overhead |\n| true | Faster pruning, reduced processing | Parents may be deleted, breaking unmined tx chains |\n\n### Recommendations\n- **false** (default) - Normal operation, preserves parent transactions\n- **true** - Skip parent preservation if you don't need to resubmit unmined transactions"`
 	SkipParentUpdates               bool          `key:"pruner_skipParentUpdates" desc:"Skip parent update operations" default:"false" category:"Pruner" usage:"Skip parent update operations" type:"bool" longdesc:"### Purpose\nSkips parent update operations during pruning.\n\n### How It Works\nWhen enabled, the pruner skips parent update operations during pruning.\n\n### Recommendations\n- **false** (default) - Normal operation\n- **true** - Skip parent update operations"`
 	SkipDeletions                   bool          `key:"pruner_skipDeletions" desc:"Skip deletion operations" default:"false" category:"Pruner" usage:"Skip deletion operations" type:"bool" longdesc:"### Purpose\nSkips deletion operations during pruning.\n\n### How It Works\nWhen enabled, the pruner skips deletion operations during pruning.\n\n### Recommendations\n- **false** (default) - Normal operation\n- **true** - Skip deletion operations"`
 }

settings/settings.go

@@ -503,10 +503,11 @@ func NewSettings(alternativeContext ...string) *Settings {
 			UTXOProgressLogInterval:        getDuration("pruner_utxoProgressLogInterval", 30*time.Second, alternativeContext...), // Progress every 30s
 			UTXOPartitionQueries:           getInt("pruner_utxoPartitionQueries", 0, alternativeContext...),                      // 0 = auto-detect based on CPU cores
 			UTXOSetTTL:                     getBool("pruner_utxoSetTTL", false, alternativeContext...),                           // Use TTL instead of delete (false = hard delete)
-			BlobDeletionEnabled:            getBool("pruner_blobDeletionEnabled", true, alternativeContext...),                   // Enable blob deletion by default
+			SkipBlobDeletion:               getBool("pruner_skipBlobDeletion", false, alternativeContext...),                     // Skip blob deletion disabled by default (deletion enabled)
 			BlobDeletionSafetyWindow:       getUint32("pruner_blobDeletionSafetyWindow", 10, alternativeContext...),              // Wait 10 blocks after persister
 			BlobDeletionBatchSize:          getInt("pruner_blobDeletionBatchSize", 1000, alternativeContext...),                  // Process 1000 deletions per batch
 			BlobDeletionMaxRetries:         getInt("pruner_blobDeletionMaxRetries", 3, alternativeContext...),                    // Retry failed deletions up to 3 times
+			SkipPreserveParents:            getBool("pruner_skipPreserveParents", false, alternativeContext...),                  // Skip Phase 1: preserve parents
 			SkipParentUpdates:              getBool("pruner_skipParentUpdates", false, alternativeContext...),                    // Skip parent updates for performance
 			SkipDeletions:                  getBool("pruner_skipDeletions", false, alternativeContext...),                        // Skip deletions for performance
 		},