@aztec/validator-ha-signer 0.0.1-commit.96bb3f7 → 0.0.1-commit.993d52e

package/src/factory.ts

@@ -3,7 +3,7 @@
  */
 import { Pool } from 'pg';
 
-import type { CreateHASignerConfig } from './config.js';
+import type { ValidatorHASignerConfig } from './config.js';
 import { PostgresSlashingProtectionDatabase } from './db/postgres.js';
 import type { CreateHASignerDeps, SlashingProtectionDatabase } from './types.js';
 import { ValidatorHASigner } from './validator_ha_signer.js';
@@ -23,7 +23,7 @@ import { ValidatorHASigner } from './validator_ha_signer.js';
  * ```typescript
  * const { signer, db } = await createHASigner({
  *   databaseUrl: process.env.DATABASE_URL,
- *   enabled: true,
+ *   haSigningEnabled: true,
  *   nodeId: 'validator-node-1',
  *   pollingIntervalMs: 100,
  *   signingTimeoutMs: 3000,
@@ -35,23 +35,15 @@ import { ValidatorHASigner } from './validator_ha_signer.js';
  * await signer.stop(); // On shutdown
  * ```
  *
- * Example with automatic migrations (simpler for dev/testing):
- * ```typescript
- * const { signer, db } = await createHASigner({
- *   databaseUrl: process.env.DATABASE_URL,
- *   enabled: true,
- *   nodeId: 'validator-node-1',
- *   runMigrations: true, // Auto-run migrations on startup
- * });
- * signer.start();
- * ```
+ * Note: Migrations must be run separately using `aztec migrate-ha-db up` before
+ * creating the signer. The factory will verify the schema is initialized via `db.initialize()`.
  *
  * @param config - Configuration for the HA signer
  * @param deps - Optional dependencies (e.g., for testing)
  * @returns An object containing the signer and database instances
  */
 export async function createHASigner(
-  config: CreateHASignerConfig,
+  config: ValidatorHASignerConfig,
   deps?: CreateHASignerDeps,
 ): Promise<{
   signer: ValidatorHASigner;
@@ -60,6 +52,9 @@ export async function createHASigner(
   const { databaseUrl, poolMaxCount, poolMinCount, poolIdleTimeoutMs, poolConnectionTimeoutMs, ...signerConfig } =
     config;
 
+  if (!databaseUrl) {
+    throw new Error('databaseUrl is required for createHASigner');
+  }
   // Create connection pool (or use provided pool)
   let pool: Pool;
   if (!deps?.pool) {

package/src/migrations.ts

@@ -3,6 +3,7 @@
  */
 import { createLogger } from '@aztec/foundation/log';
 
+import { readdirSync } from 'fs';
 import { runner } from 'node-pg-migrate';
 import { dirname, join } from 'path';
 import { fileURLToPath } from 'url';
@@ -30,17 +31,32 @@ export async function runMigrations(databaseUrl: string, options: RunMigrationsO
 
   const log = createLogger('validator-ha-signer:migrations');
 
+  const migrationsDir = join(__dirname, 'db', 'migrations');
+
   try {
     log.info(`Running migrations ${direction}...`);
 
+    // Filter out .d.ts and .d.ts.map files - node-pg-migrate only needs .js files
+    const migrationFiles = readdirSync(migrationsDir);
+    const jsMigrationFiles = migrationFiles.filter(
+      file => file.endsWith('.js') && !file.endsWith('.d.ts') && !file.endsWith('.d.ts.map'),
+    );
+
+    if (jsMigrationFiles.length === 0) {
+      log.info('No migration files found');
+      return [];
+    }
+
     const appliedMigrations = await runner({
       databaseUrl,
-      dir: join(__dirname, 'db', 'migrations'),
+      dir: migrationsDir,
       direction,
       migrationsTable: 'pgmigrations',
       count: direction === 'down' ? 1 : Infinity,
       verbose,
       log: msg => (verbose ? log.info(msg) : log.debug(msg)),
+      // Ignore TypeScript declaration files - node-pg-migrate will try to import them otherwise
+      ignorePattern: '.*\\.d\\.(ts|js)$|.*\\.d\\.ts\\.map$',
     });
 
     if (appliedMigrations.length === 0) {

package/src/slashing_protection_service.ts

@@ -8,9 +8,15 @@ import { type Logger, createLogger } from '@aztec/foundation/log';
 import { RunningPromise } from '@aztec/foundation/promise';
 import { sleep } from '@aztec/foundation/sleep';
 
-import { type CheckAndRecordParams, type DeleteDutyParams, DutyStatus, type RecordSuccessParams } from './db/types.js';
+import {
+  type CheckAndRecordParams,
+  type DeleteDutyParams,
+  DutyStatus,
+  type RecordSuccessParams,
+  getBlockIndexFromDutyIdentifier,
+} from './db/types.js';
 import { DutyAlreadySignedError, SlashingProtectionError } from './errors.js';
-import type { SlashingProtectionConfig, SlashingProtectionDatabase } from './types.js';
+import type { SlashingProtectionDatabase, ValidatorHASignerConfig } from './types.js';
 
 /**
  * Slashing Protection Service
@@ -31,22 +37,22 @@ export class SlashingProtectionService {
   private readonly log: Logger;
   private readonly pollingIntervalMs: number;
   private readonly signingTimeoutMs: number;
+  private readonly maxStuckDutiesAgeMs: number;
 
   private cleanupRunningPromise: RunningPromise;
+  private lastOldDutiesCleanupAtMs?: number;
 
   constructor(
     private readonly db: SlashingProtectionDatabase,
-    private readonly config: SlashingProtectionConfig,
+    private readonly config: ValidatorHASignerConfig,
   ) {
     this.log = createLogger('slashing-protection');
     this.pollingIntervalMs = config.pollingIntervalMs;
     this.signingTimeoutMs = config.signingTimeoutMs;
+    // Default to 144s (2x 72s Aztec slot duration) if not explicitly configured
+    this.maxStuckDutiesAgeMs = config.maxStuckDutiesAgeMs ?? 144_000;
 
-    this.cleanupRunningPromise = new RunningPromise(
-      this.cleanupStuckDuties.bind(this),
-      this.log,
-      this.config.maxStuckDutiesAgeMs,
-    );
+    this.cleanupRunningPromise = new RunningPromise(this.cleanup.bind(this), this.log, this.maxStuckDutiesAgeMs);
   }
 
   /**
@@ -58,7 +64,6 @@ export class SlashingProtectionService {
    * 2. If insert succeeds, we acquired the lock - return the lockToken
    * 3. If a record exists, handle based on status:
    *    - SIGNED: Throw appropriate error (already signed or slashing protection)
-   *    - FAILED: Delete the failed record
    *    - SIGNING: Wait and poll until status changes, then handle result
    *
    * @returns The lockToken that must be used for recordSuccess/deleteDuty
@@ -98,9 +103,16 @@ export class SlashingProtectionService {
             existingNodeId: record.nodeId,
             attemptingNodeId: nodeId,
           });
-          throw new SlashingProtectionError(slot, dutyType, record.messageHash, messageHash);
+          throw new SlashingProtectionError(
+            slot,
+            dutyType,
+            record.blockIndexWithinCheckpoint,
+            record.messageHash,
+            messageHash,
+            record.nodeId,
+          );
         }
-        throw new DutyAlreadySignedError(slot, dutyType, record.nodeId);
+        throw new DutyAlreadySignedError(slot, dutyType, record.blockIndexWithinCheckpoint, record.nodeId);
       } else if (record.status === DutyStatus.SIGNING) {
         // Another node is currently signing - check for timeout
         if (Date.now() - startTime > this.signingTimeoutMs) {
@@ -109,7 +121,7 @@ export class SlashingProtectionService {
             timeoutMs: this.signingTimeoutMs,
             signingNodeId: record.nodeId,
           });
-          throw new DutyAlreadySignedError(slot, dutyType, 'unknown (timeout)');
+          throw new DutyAlreadySignedError(slot, dutyType, record.blockIndexWithinCheckpoint, 'unknown (timeout)');
         }
 
         // Wait and poll
@@ -133,9 +145,18 @@ export class SlashingProtectionService {
    * @returns true if the update succeeded, false if token didn't match
    */
   async recordSuccess(params: RecordSuccessParams): Promise<boolean> {
-    const { validatorAddress, slot, dutyType, signature, nodeId, lockToken } = params;
+    const { rollupAddress, validatorAddress, slot, dutyType, signature, nodeId, lockToken } = params;
+    const blockIndexWithinCheckpoint = getBlockIndexFromDutyIdentifier(params);
 
-    const success = await this.db.updateDutySigned(validatorAddress, slot, dutyType, signature.toString(), lockToken);
+    const success = await this.db.updateDutySigned(
+      rollupAddress,
+      validatorAddress,
+      slot,
+      dutyType,
+      signature.toString(),
+      lockToken,
+      blockIndexWithinCheckpoint,
+    );
 
     if (success) {
       this.log.info(`Recorded successful signing for duty ${dutyType} at slot ${slot}`, {
@@ -160,9 +181,17 @@ export class SlashingProtectionService {
    * @returns true if the delete succeeded, false if token didn't match
    */
   async deleteDuty(params: DeleteDutyParams): Promise<boolean> {
-    const { validatorAddress, slot, dutyType, lockToken } = params;
+    const { rollupAddress, validatorAddress, slot, dutyType, lockToken } = params;
+    const blockIndexWithinCheckpoint = getBlockIndexFromDutyIdentifier(params);
 
-    const success = await this.db.deleteDuty(validatorAddress, slot, dutyType, lockToken);
+    const success = await this.db.deleteDuty(
+      rollupAddress,
+      validatorAddress,
+      slot,
+      dutyType,
+      lockToken,
+      blockIndexWithinCheckpoint,
+    );
 
     if (success) {
       this.log.info(`Deleted duty ${dutyType} at slot ${slot} to allow retry`, {
@@ -188,7 +217,19 @@ export class SlashingProtectionService {
    * Start running tasks.
    * Cleanup runs immediately on start to recover from any previous crashes.
    */
-  start() {
+  /**
+   * Start the background cleanup task.
+   * Also performs one-time cleanup of duties with outdated rollup addresses.
+   */
+  async start() {
+    // One-time cleanup at startup: remove duties from previous rollup versions
+    const numOutdatedRollupDuties = await this.db.cleanupOutdatedRollupDuties(this.config.l1Contracts.rollupAddress);
+    if (numOutdatedRollupDuties > 0) {
+      this.log.info(`Cleaned up ${numOutdatedRollupDuties} duties with outdated rollup address at startup`, {
+        currentRollupAddress: this.config.l1Contracts.rollupAddress.toString(),
+      });
+    }
+
     this.cleanupRunningPromise.start();
     this.log.info('Slashing protection service started', { nodeId: this.config.nodeId });
   }
@@ -202,15 +243,45 @@ export class SlashingProtectionService {
   }
 
   /**
-   * Cleanup own stuck duties
+   * Close the database connection.
+   * Should be called after stop() during graceful shutdown.
    */
-  private async cleanupStuckDuties() {
-    const numDuties = await this.db.cleanupOwnStuckDuties(this.config.nodeId, this.config.maxStuckDutiesAgeMs);
-    if (numDuties > 0) {
-      this.log.info(`Cleaned up ${numDuties} stuck duties`, {
+  async close() {
+    await this.db.close();
+    this.log.info('Slashing protection database connection closed');
+  }
+
+  /**
+   * Periodic cleanup of stuck duties and optionally old signed duties.
+   * Runs in the background via RunningPromise.
+   */
+  private async cleanup() {
+    // 1. Clean up stuck duties (our own node's duties that got stuck in 'signing' status)
+    const numStuckDuties = await this.db.cleanupOwnStuckDuties(this.config.nodeId, this.maxStuckDutiesAgeMs);
+    if (numStuckDuties > 0) {
+      this.log.verbose(`Cleaned up ${numStuckDuties} stuck duties`, {
         nodeId: this.config.nodeId,
-        maxStuckDutiesAgeMs: this.config.maxStuckDutiesAgeMs,
+        maxStuckDutiesAgeMs: this.maxStuckDutiesAgeMs,
       });
     }
+
+    // 2. Clean up old signed duties if configured
+    // we shouldn't run this as often as stuck duty cleanup.
+    if (this.config.cleanupOldDutiesAfterHours !== undefined) {
+      const maxAgeMs = this.config.cleanupOldDutiesAfterHours * 60 * 60 * 1000;
+      const nowMs = Date.now();
+      const shouldRun =
+        this.lastOldDutiesCleanupAtMs === undefined || nowMs - this.lastOldDutiesCleanupAtMs >= maxAgeMs;
+      if (shouldRun) {
+        const numOldDuties = await this.db.cleanupOldDuties(maxAgeMs);
+        this.lastOldDutiesCleanupAtMs = nowMs;
+        if (numOldDuties > 0) {
+          this.log.verbose(`Cleaned up ${numOldDuties} old signed duties`, {
+            cleanupOldDutiesAfterHours: this.config.cleanupOldDutiesAfterHours,
+            maxAgeMs,
+          });
+        }
+      }
+    }
   }
 }

package/src/test/pglite_pool.ts

@@ -0,0 +1,256 @@
+/**
+ * Vendored pg-compatible Pool/Client wrapper for PGlite.
+ *
+ * Copied from @middle-management/pglite-pg-adapter v0.0.3
+ * https://www.npmjs.com/package/@middle-management/pglite-pg-adapter
+ *
+ * Modifications:
+ * - Converted to ESM and TypeScript
+ * - Uses PGliteInterface instead of PGlite class to avoid TypeScript
+ *   type mismatches from ESM/CJS dual package resolution with private fields
+ * - Simplified rowCount calculation to handle CTEs properly
+ */
+import type { PGliteInterface } from '@electric-sql/pglite';
+import { EventEmitter } from 'events';
+import type { QueryResult, QueryResultRow } from 'pg';
+import { Readable, Writable } from 'stream';
+
+export interface PoolConfig {
+  pglite: PGliteInterface;
+  max?: number;
+  min?: number;
+}
+
+interface ClientConfig {
+  pglite: PGliteInterface;
+  host?: string;
+  port?: number;
+  ssl?: boolean;
+}
+
+export class Client extends EventEmitter {
+  protected pglite: PGliteInterface;
+  protected _connected = false;
+  readonly host: string;
+  readonly port: number;
+  readonly ssl: boolean;
+  readonly connection: object;
+
+  // Stub implementations for pg compatibility
+  readonly copyFrom = (): Writable => new Writable();
+  readonly copyTo = (): Readable => new Readable();
+  readonly pauseDrain = (): void => {};
+  readonly resumeDrain = (): void => {};
+  readonly escapeLiteral = (str: string): string => `'${str.replace(/'/g, "''")}'`;
+  readonly escapeIdentifier = (str: string): string => `"${str.replace(/"/g, '""')}"`;
+  readonly setTypeParser = (): void => {};
+  readonly getTypeParser = (): ((value: string) => unknown) => (value: string) => value;
+
+  constructor(config: ClientConfig) {
+    super();
+    this.pglite = config.pglite;
+    this.host = config.host || 'localhost';
+    this.port = config.port || 5432;
+    this.ssl = typeof config.ssl === 'boolean' ? config.ssl : !!config.ssl;
+    this.connection = {};
+  }
+
+  connect(): Promise<void> {
+    if (this._connected) {
+      return Promise.resolve();
+    }
+    this._connected = true;
+    this.emit('connect');
+    return Promise.resolve();
+  }
+
+  end(): Promise<void> {
+    if (!this._connected) {
+      return Promise.resolve();
+    }
+    this._connected = false;
+    this.emit('end');
+    return Promise.resolve();
+  }
+
+  async query<R extends QueryResultRow = any>(text: string, values?: any[]): Promise<QueryResult<R>> {
+    if (!this._connected) {
+      throw new Error('Client is not connected');
+    }
+    const result = await this.pglite.query<R>(text, values);
+    return this.convertPGliteResult(result);
+  }
+
+  protected convertPGliteResult<R extends QueryResultRow>(result: {
+    rows: R[];
+    fields: Array<{ name: string; dataTypeID: number }>;
+    affectedRows?: number;
+  }): QueryResult<R> {
+    return {
+      command: '',
+      rowCount: 'affectedRows' in result ? (result.affectedRows ?? 0) : result.rows.length,
+      oid: 0,
+      fields: result.fields.map(field => ({
+        name: field.name,
+        tableID: 0,
+        columnID: 0,
+        dataTypeID: field.dataTypeID,
+        dataTypeSize: -1,
+        dataTypeModifier: -1,
+        format: 'text',
+      })),
+      rows: result.rows,
+    };
+  }
+
+  get connected(): boolean {
+    return this._connected;
+  }
+}
+
+export class Pool extends EventEmitter {
+  private clients: PoolClient[] = [];
+  private availableClients: PoolClient[] = [];
+  private waitingQueue: Array<(client: PoolClient) => void> = [];
+  private _ended = false;
+  private pglite: PGliteInterface;
+  private _config: PoolConfig;
+
+  readonly expiredCount = 0;
+  readonly options: PoolConfig;
+
+  constructor(config: PoolConfig) {
+    super();
+    this._config = { max: 10, min: 0, ...config };
+    this.pglite = config.pglite;
+    this.options = config;
+  }
+
+  get totalCount(): number {
+    return this.clients.length;
+  }
+
+  get idleCount(): number {
+    return this.availableClients.length;
+  }
+
+  get waitingCount(): number {
+    return this.waitingQueue.length;
+  }
+
+  get ending(): boolean {
+    return this._ended;
+  }
+
+  get ended(): boolean {
+    return this._ended;
+  }
+
+  connect(): Promise<PoolClient> {
+    if (this._ended) {
+      return Promise.reject(new Error('Pool is ended'));
+    }
+
+    if (this.availableClients.length > 0) {
+      const client = this.availableClients.pop()!;
+      client._markInUse();
+      return Promise.resolve(client);
+    }
+
+    if (this.clients.length < (this._config.max || 10)) {
+      const client = new PoolClient(this.pglite, this);
+      this.clients.push(client);
+      return Promise.resolve(client);
+    }
+
+    return new Promise(resolve => {
+      this.waitingQueue.push(resolve);
+    });
+  }
+
+  async query<R extends QueryResultRow = any>(text: string, values?: any[]): Promise<QueryResult<R>> {
+    const client = await this.connect();
+    try {
+      return await client.query<R>(text, values);
+    } finally {
+      client.release();
+    }
+  }
+
+  releaseClient(client: PoolClient): void {
+    const index = this.clients.indexOf(client);
+    if (index !== -1) {
+      client._markAvailable();
+      if (this.waitingQueue.length > 0) {
+        const resolve = this.waitingQueue.shift()!;
+        client._markInUse();
+        resolve(client);
+      } else {
+        this.availableClients.push(client);
+      }
+    }
+  }
+
+  end(): Promise<void> {
+    this._ended = true;
+    this.clients.forEach(client => client._markReleased());
+    this.clients = [];
+    this.availableClients = [];
+    this.emit('end');
+    return Promise.resolve();
+  }
+}
+
+export class PoolClient extends Client {
+  private pool: Pool;
+  private _released = false;
+  private _inUse = true;
+  private _userReleased = false;
+
+  constructor(pglite: PGliteInterface, pool: Pool) {
+    super({ pglite });
+    this.pool = pool;
+    this._connected = true;
+  }
+
+  override async query<R extends QueryResultRow = any>(text: string, values?: any[]): Promise<QueryResult<R>> {
+    if (this._userReleased && !this._inUse) {
+      throw new Error('Client has been released back to the pool');
+    }
+    const result = await this.pglite.query<R>(text, values);
+    return this.convertPGliteResult(result);
+  }
+
+  release(): void {
+    if (this._released || this._userReleased) {
+      return;
+    }
+    this._userReleased = true;
+    this.pool.releaseClient(this);
+  }
+
+  override end(): Promise<void> {
+    this.release();
+    return Promise.resolve();
+  }
+
+  _markInUse(): void {
+    this._inUse = true;
+    this._userReleased = false;
+  }
+
+  _markAvailable(): void {
+    this._inUse = false;
+    this._userReleased = false;
+  }
+
+  _markReleased(): void {
+    this._released = true;
+    this._inUse = false;
+    this._userReleased = true;
+  }
+
+  override get connected(): boolean {
+    return this._connected && !this._released;
+  }
+}