feat: track world state metrics

yarn-project/aztec-node/src/aztec-node/server.ts

@@ -152,7 +152,7 @@ export class AztecNodeService implements AztecNode {
     );
 
     // now create the merkle trees and the world state synchronizer
-    const worldStateSynchronizer = await createWorldStateSynchronizer(config, store, archiver);
+    const worldStateSynchronizer = await createWorldStateSynchronizer(config, store, archiver, telemetry);
 
     // start both and wait for them to sync from the block source
     await Promise.all([p2pClient.start(), worldStateSynchronizer.start()]);
@@ -722,7 +722,7 @@ export class AztecNodeService implements AztecNode {
     // Instantiate merkle trees so uncommitted updates by this simulation are local to it.
     // TODO we should be able to remove this after https://github.com/AztecProtocol/aztec-packages/issues/1869
     // So simulation of public functions doesn't affect the merkle trees.
-    const merkleTrees = await MerkleTrees.new(this.merkleTreesDb, this.log);
+    const merkleTrees = await MerkleTrees.new(this.merkleTreesDb, new NoopTelemetryClient(), this.log);
 
     const publicProcessorFactory = new PublicProcessorFactory(
       merkleTrees.asLatest(),

yarn-project/end-to-end/src/composed/integration_l1_publisher.test.ts

@@ -148,7 +148,7 @@ describe('L1Publisher integration', () => {
     });
 
     const tmpStore = openTmpStore();
-    builderDb = await MerkleTrees.new(tmpStore);
+    builderDb = await MerkleTrees.new(tmpStore, new NoopTelemetryClient());
     blockSource = mock<ArchiveSource>();
     blockSource.getBlocks.mockResolvedValue([]);
     const worldStateConfig: WorldStateConfig = {

yarn-project/kv-store/src/interfaces/store.ts

@@ -68,4 +68,9 @@ export interface AztecKVStore {
    * Deletes the store
    */
   delete(): Promise<void>;
+
+  /**
+   * Estimates the size of the store in bytes.
+   */
+  estimateSize(): { bytes: number };
 }

yarn-project/kv-store/src/lmdb/store.ts

@@ -154,4 +154,16 @@ export class AztecLmdbStore implements AztecKVStore {
   async delete() {
     await this.#rootDb.drop();
   }
+
+  estimateSize(): { bytes: number } {
+    const stats = this.#rootDb.getStats();
+    // `mapSize` represents to total amount of memory currently being used by the database.
+    // since the database is mmap'd, this is a good estimate of the size of the database for now.
+    // http://www.lmdb.tech/doc/group__mdb.html#a4bde3c8b676457342cba2fe27aed5fbd
+    if ('mapSize' in stats && typeof stats.mapSize === 'number') {
+      return { bytes: stats.mapSize };
+    } else {
+      return { bytes: 0 };
+    }
+  }
 }

yarn-project/prover-client/src/mocks/test_context.ts

@@ -68,8 +68,8 @@ export class TestContext {
     const publicContractsDB = mock<ContractsDataSourcePublicDB>();
     const publicWorldStateDB = mock<WorldStatePublicDB>();
     const publicKernel = new RealPublicKernelCircuitSimulator(new WASMSimulator());
-    const actualDb = await MerkleTrees.new(openTmpStore()).then(t => t.asLatest());
     const telemetry = new NoopTelemetryClient();
+    const actualDb = await MerkleTrees.new(openTmpStore(), telemetry).then(t => t.asLatest());
     const processor = new PublicProcessor(
       actualDb,
       publicExecutor,

yarn-project/prover-client/src/orchestrator/orchestrator_mixed_blocks_2.test.ts

@@ -5,6 +5,7 @@ import { range } from '@aztec/foundation/array';
 import { times } from '@aztec/foundation/collection';
 import { createDebugLogger } from '@aztec/foundation/log';
 import { openTmpStore } from '@aztec/kv-store/utils';
+import { NoopTelemetryClient } from '@aztec/telemetry-client/noop';
 import { type MerkleTreeOperations, MerkleTrees } from '@aztec/world-state';
 
 import { makeBloatedProcessedTx, updateExpectedTreesFromTxs } from '../mocks/fixtures.js';
@@ -18,7 +19,7 @@ describe('prover/orchestrator/mixed-blocks', () => {
 
   beforeEach(async () => {
     context = await TestContext.new(logger);
-    expectsDb = await MerkleTrees.new(openTmpStore()).then(t => t.asLatest());
+    expectsDb = await MerkleTrees.new(openTmpStore(), new NoopTelemetryClient()).then(t => t.asLatest());
   });
 
   afterEach(async () => {

yarn-project/prover-client/src/orchestrator/orchestrator_single_blocks.test.ts

@@ -5,6 +5,7 @@ import { range } from '@aztec/foundation/array';
 import { createDebugLogger } from '@aztec/foundation/log';
 import { sleep } from '@aztec/foundation/sleep';
 import { openTmpStore } from '@aztec/kv-store/utils';
+import { NoopTelemetryClient } from '@aztec/telemetry-client/noop';
 import { type MerkleTreeOperations, MerkleTrees } from '@aztec/world-state';
 
 import { makeBloatedProcessedTx, updateExpectedTreesFromTxs } from '../mocks/fixtures.js';
@@ -18,7 +19,7 @@ describe('prover/orchestrator/blocks', () => {
 
   beforeEach(async () => {
     context = await TestContext.new(logger);
-    expectsDb = await MerkleTrees.new(openTmpStore()).then(t => t.asLatest());
+    expectsDb = await MerkleTrees.new(openTmpStore(), new NoopTelemetryClient()).then(t => t.asLatest());
   });
 
   afterEach(async () => {

yarn-project/prover-client/src/orchestrator/orchestrator_workflow.test.ts

@@ -23,9 +23,10 @@ describe('prover/orchestrator', () => {
     let mockProver: MockProxy<ServerCircuitProver>;
     let actualDb: MerkleTreeOperations;
     beforeEach(async () => {
-      actualDb = await MerkleTrees.new(openTmpStore()).then(t => t.asLatest());
+      const telemetryClient = new NoopTelemetryClient();
+      actualDb = await MerkleTrees.new(openTmpStore(), telemetryClient).then(t => t.asLatest());
       mockProver = mock<ServerCircuitProver>();
-      orchestrator = new ProvingOrchestrator(actualDb, mockProver, new NoopTelemetryClient());
+      orchestrator = new ProvingOrchestrator(actualDb, mockProver, telemetryClient);
     });
 
     it('calls root parity circuit only when ready', async () => {

yarn-project/prover-node/src/factory.ts

@@ -35,7 +35,7 @@ export async function createProverNode(
   log.verbose(`Created archiver and synced to block ${await archiver.getBlockNumber()}`);
 
   const worldStateConfig = { ...config, worldStateProvenBlocksOnly: true };
-  const worldStateSynchronizer = await createWorldStateSynchronizer(worldStateConfig, store, archiver);
+  const worldStateSynchronizer = await createWorldStateSynchronizer(worldStateConfig, store, archiver, telemetry);
   await worldStateSynchronizer.start();
 
   const simulationProvider = await createSimulationProvider(config, log);

yarn-project/telemetry-client/src/attributes.ts

@@ -60,3 +60,5 @@ export const L1_TX_TYPE = 'aztec.l1.tx_type';
 export const TX_PHASE_NAME = 'aztec.tx.phase_name';
 /** The proving job type */
 export const PROVING_JOB_TYPE = 'aztec.proving.job_type';
+
+export const MERKLE_TREE_NAME = 'aztec.merkle_tree.name';

yarn-project/telemetry-client/src/metrics.ts

@@ -62,3 +62,8 @@ export const PROVING_ORCHESTRATOR_BASE_ROLLUP_INPUTS_DURATION =
 
 export const PROVING_QUEUE_JOB_SIZE = 'aztec.proving_queue.job_size';
 export const PROVING_QUEUE_SIZE = 'aztec.proving_queue.size';
+
+export const WORLD_STATE_FORK_DURATION = 'aztec.world_state.fork.duration';
+export const WORLD_STATE_SYNC_DURATION = 'aztec.world_state.sync.duration';
+export const WORLD_STATE_MERKLE_TREE_SIZE = 'aztec.world_state.merkle_tree_size';
+export const WORLD_STATE_DB_SIZE = 'aztec.world_state.db_size';

yarn-project/txe/src/txe_service/txe_service.ts

@@ -18,6 +18,7 @@ import { type Logger } from '@aztec/foundation/log';
 import { KeyStore } from '@aztec/key-store';
 import { openTmpStore } from '@aztec/kv-store/utils';
 import { ExecutionNoteCache, PackedValuesCache, type TypedOracle } from '@aztec/simulator';
+import { NoopTelemetryClient } from '@aztec/telemetry-client/noop';
 import { MerkleTrees } from '@aztec/world-state';
 
 import { TXE } from '../oracle/txe_oracle.js';
@@ -38,7 +39,7 @@ export class TXEService {
 
   static async init(logger: Logger) {
     const store = openTmpStore(true);
-    const trees = await MerkleTrees.new(store, logger);
+    const trees = await MerkleTrees.new(store, new NoopTelemetryClient(), logger);
     const packedValuesCache = new PackedValuesCache();
     const txHash = new Fr(1); // The txHash is used for computing the revertible nullifiers for non-revertible note hashes. It can be any value for testing.
     const noteCache = new ExecutionNoteCache(txHash);

yarn-project/world-state/package.json

@@ -58,6 +58,7 @@
     "@aztec/foundation": "workspace:^",
     "@aztec/kv-store": "workspace:^",
     "@aztec/merkle-tree": "workspace:^",
+    "@aztec/telemetry-client": "workspace:^",
     "@aztec/types": "workspace:^",
     "tslib": "^2.4.0"
   },

yarn-project/world-state/src/synchronizer/factory.ts

@@ -1,5 +1,6 @@
 import { type L1ToL2MessageSource, type L2BlockSource } from '@aztec/circuit-types';
 import { type AztecKVStore } from '@aztec/kv-store';
+import { type TelemetryClient } from '@aztec/telemetry-client';
 
 import { MerkleTrees } from '../world-state-db/merkle_trees.js';
 import { type WorldStateConfig } from './config.js';
@@ -9,7 +10,8 @@ export async function createWorldStateSynchronizer(
   config: WorldStateConfig,
   store: AztecKVStore,
   l2BlockSource: L2BlockSource & L1ToL2MessageSource,
+  client: TelemetryClient,
 ) {
-  const merkleTrees = await MerkleTrees.new(store);
+  const merkleTrees = await MerkleTrees.new(store, client);
   return new ServerWorldStateSynchronizer(store, merkleTrees, l2BlockSource, config);
 }

yarn-project/world-state/src/world-state-db/merkle_trees.ts

@@ -32,6 +32,7 @@ import {
 import { padArrayEnd } from '@aztec/foundation/collection';
 import { type DebugLogger, createDebugLogger } from '@aztec/foundation/log';
 import { SerialQueue } from '@aztec/foundation/queue';
+import { Timer, elapsed } from '@aztec/foundation/timer';
 import { type IndexedTreeLeafPreimage } from '@aztec/foundation/trees';
 import { type AztecKVStore, type AztecSingleton } from '@aztec/kv-store';
 import {
@@ -45,6 +46,7 @@ import {
   loadTree,
   newTree,
 } from '@aztec/merkle-tree';
+import { type TelemetryClient } from '@aztec/telemetry-client';
 import { type Hasher } from '@aztec/types/interfaces';
 
 import {
@@ -55,6 +57,7 @@ import {
 } from './merkle_tree_db.js';
 import { type MerkleTreeMap } from './merkle_tree_map.js';
 import { MerkleTreeOperationsFacade } from './merkle_tree_operations_facade.js';
+import { WorldStateMetrics } from './metrics.js';
 
 /**
  * The nullifier tree is an indexed tree.
@@ -98,18 +101,20 @@ export class MerkleTrees implements MerkleTreeDb {
   private trees: MerkleTreeMap = null as any;
   private jobQueue = new SerialQueue();
   private initialStateReference: AztecSingleton<Buffer>;
+  private metrics: WorldStateMetrics;
 
-  private constructor(private store: AztecKVStore, private log: DebugLogger) {
+  private constructor(private store: AztecKVStore, private telemetryClient: TelemetryClient, private log: DebugLogger) {
     this.initialStateReference = store.openSingleton('merkle_trees_initial_state_reference');
+    this.metrics = new WorldStateMetrics(telemetryClient);
   }
 
   /**
    * Method to asynchronously create and initialize a MerkleTrees instance.
    * @param store - The db instance to use for data persistance.
    * @returns - A fully initialized MerkleTrees instance.
    */
-  public static async new(store: AztecKVStore, log = createDebugLogger('aztec:merkle_trees')) {
-    const merkleTrees = new MerkleTrees(store, log);
+  public static async new(store: AztecKVStore, client: TelemetryClient, log = createDebugLogger('aztec:merkle_trees')) {
+    const merkleTrees = new MerkleTrees(store, client, log);
     await merkleTrees.#init();
     return merkleTrees;
   }
@@ -181,12 +186,17 @@ export class MerkleTrees implements MerkleTreeDb {
   }
 
   public async fork(): Promise<MerkleTrees> {
-    // TODO(palla/prover-node): If the underlying store is being shared with other components, we're unnecessarily
-    // copying a lot of data unrelated to merkle trees. This may be fine for now, and we may be able to ditch backup-based
-    // forking in favor of a more elegant proposal. But if we see this operation starts taking a lot of time, we may want
-    // to open separate stores for merkle trees and other components.
-    const forked = await this.store.fork();
-    return MerkleTrees.new(forked, this.log);
+    const [ms, db] = await elapsed(async () => {
+      // TODO(palla/prover-node): If the underlying store is being shared with other components, we're unnecessarily
+      // copying a lot of data unrelated to merkle trees. This may be fine for now, and we may be able to ditch backup-based
+      // forking in favor of a more elegant proposal. But if we see this operation starts taking a lot of time, we may want
+      // to open separate stores for merkle trees and other components.
+      const forked = await this.store.fork();
+      return MerkleTrees.new(forked, this.telemetryClient, this.log);
+    });
+
+    this.metrics.recordForkDuration(ms);
+    return db;
   }
 
   public async delete() {
@@ -581,6 +591,8 @@ export class MerkleTrees implements MerkleTreeDb {
    * @param l1ToL2Messages - The L1 to L2 messages for the block.
    */
   async #handleL2BlockAndMessages(l2Block: L2Block, l1ToL2Messages: Fr[]): Promise<HandleL2BlockAndMessagesResult> {
+    const timer = new Timer();
+
     const treeRootWithIdPairs = [
       [l2Block.header.state.partial.nullifierTree.root, MerkleTreeId.NULLIFIER_TREE],
       [l2Block.header.state.partial.noteHashTree.root, MerkleTreeId.NOTE_HASH_TREE],
@@ -664,10 +676,13 @@ export class MerkleTrees implements MerkleTreeDb {
         );
       } else {
         this.log.debug(`Tree ${treeName} synched with size ${info.size} root ${rootStr}`);
+        this.metrics.recordTreeSize(treeName, info.size);
       }
     }
     await this.#snapshot(l2Block.number);
 
+    this.metrics.recordDbSize(this.store.estimateSize().bytes);
+    this.metrics.recordSyncDuration(ourBlock ? 'commit' : 'rollback_and_update', timer);
     return { isBlockOurs: ourBlock };
   }
 

yarn-project/world-state/src/world-state-db/metrics.ts

@@ -0,0 +1,64 @@
+import { type Timer } from '@aztec/foundation/timer';
+import {
+  Attributes,
+  type Gauge,
+  type Histogram,
+  Metrics,
+  type TelemetryClient,
+  ValueType,
+} from '@aztec/telemetry-client';
+
+export class WorldStateMetrics {
+  private treeSize: Gauge;
+  private dbSize: Gauge;
+  private forkDuration: Histogram;
+  private syncDuration: Histogram;
+
+  constructor(client: TelemetryClient, name = 'MerkleTreesDb') {
+    const meter = client.getMeter(name);
+    this.treeSize = meter.createGauge(Metrics.WORLD_STATE_MERKLE_TREE_SIZE, {
+      description: 'The size of Merkle trees',
+      valueType: ValueType.INT,
+    });
+
+    this.dbSize = meter.createGauge(Metrics.WORLD_STATE_DB_SIZE, {
+      description: 'The size of the World State DB',
+      valueType: ValueType.INT,
+      unit: 'By',
+    });
+
+    this.forkDuration = meter.createHistogram(Metrics.WORLD_STATE_FORK_DURATION, {
+      description: 'The duration of a fork operation',
+      unit: 'ms',
+      valueType: ValueType.INT,
+    });
+
+    this.syncDuration = meter.createHistogram(Metrics.WORLD_STATE_SYNC_DURATION, {
+      description: 'The duration of a sync operation',
+      unit: 'ms',
+      valueType: ValueType.INT,
+    });
+  }
+
+  recordTreeSize(treeName: string, treeSize: bigint) {
+    this.treeSize.record(Number(treeSize), {
+      [Attributes.MERKLE_TREE_NAME]: treeName,
+    });
+  }
+
+  recordDbSize(dbSizeInBytes: number) {
+    this.dbSize.record(dbSizeInBytes);
+  }
+
+  recordForkDuration(timerOrMs: Timer | number) {
+    const ms = Math.ceil(typeof timerOrMs === 'number' ? timerOrMs : timerOrMs.ms());
+    this.forkDuration.record(ms);
+  }
+
+  recordSyncDuration(syncType: 'commit' | 'rollback_and_update', timerOrMs: Timer | number) {
+    const ms = Math.ceil(typeof timerOrMs === 'number' ? timerOrMs : timerOrMs.ms());
+    this.syncDuration.record(ms, {
+      [Attributes.STATUS]: syncType,
+    });
+  }
+}

yarn-project/world-state/tsconfig.json

@@ -21,6 +21,9 @@
     {
       "path": "../merkle-tree"
     },
+    {
+      "path": "../telemetry-client"
+    },
     {
       "path": "../types"
     }

yarn-project/yarn.lock

@@ -1208,6 +1208,7 @@ __metadata:
     "@aztec/foundation": "workspace:^"
     "@aztec/kv-store": "workspace:^"
     "@aztec/merkle-tree": "workspace:^"
+    "@aztec/telemetry-client": "workspace:^"
     "@aztec/types": "workspace:^"
     "@jest/globals": ^29.5.0
     "@types/jest": ^29.5.0