@lodestar/beacon-node 1.51.0-dev.02c5a19f35 → 1.51.0-dev.0586e6e617

package/src/network/peers/utils/assertPeerRelevance.ts

@@ -1,5 +1,6 @@
-import {ForkName, isForkPostFulu} from "@lodestar/params";
-import {ForkDigest, Root, Slot, Status, fulu, ssz} from "@lodestar/types";
+import {ForkName, GENESIS_EPOCH, isForkPostFulu} from "@lodestar/params";
+import {computeEpochAtSlot} from "@lodestar/state-transition";
+import {Epoch, ForkDigest, Root, Slot, Status, fulu, ssz} from "@lodestar/types";
 import {toHex, toRootHex} from "@lodestar/utils";
 
 // TODO: Why this value? (From Lighthouse)
@@ -9,6 +10,8 @@ export enum IrrelevantPeerCode {
   INCOMPATIBLE_FORKS = "IRRELEVANT_PEER_INCOMPATIBLE_FORKS",
   DIFFERENT_CLOCKS = "IRRELEVANT_PEER_DIFFERENT_CLOCKS",
   DIFFERENT_FINALIZED = "IRRELEVANT_PEER_DIFFERENT_FINALIZED",
+  FINALIZED_EPOCH_TOO_HIGH = "IRRELEVANT_PEER_FINALIZED_EPOCH_TOO_HIGH",
+  INVALID_FINALIZED_ROOT = "IRRELEVANT_PEER_INVALID_FINALIZED_ROOT",
   NO_EARLIEST_AVAILABLE_SLOT = "NO_EARLIEST_AVAILABLE_SLOT",
 }
 
@@ -16,6 +19,8 @@ type IrrelevantPeerType =
   | {code: IrrelevantPeerCode.INCOMPATIBLE_FORKS; ours: ForkDigest; theirs: ForkDigest}
   | {code: IrrelevantPeerCode.DIFFERENT_CLOCKS; slotDiff: number}
   | {code: IrrelevantPeerCode.NO_EARLIEST_AVAILABLE_SLOT}
+  | {code: IrrelevantPeerCode.FINALIZED_EPOCH_TOO_HIGH; finalizedEpoch: Epoch; maxFinalizedEpoch: Epoch}
+  | {code: IrrelevantPeerCode.INVALID_FINALIZED_ROOT}
   | {code: IrrelevantPeerCode.DIFFERENT_FINALIZED; expectedRoot: Root; remoteRoot: Root};
 
 /**
@@ -45,29 +50,35 @@ export function assertPeerRelevance(
     return {code: IrrelevantPeerCode.DIFFERENT_CLOCKS, slotDiff};
   }
 
-  // The remote's finalized epoch is less than or equal to ours, but the block root is
-  // different to the one in our chain. Therefore, the node is on a different chain and we
-  // should not communicate with them.
+  // A checkpoint can be finalized at the earliest at the start of epoch + 2
+  const maxFinalizedEpoch = Math.max(
+    GENESIS_EPOCH,
+    computeEpochAtSlot(Math.max(currentSlot, 0) + FUTURE_SLOT_TOLERANCE) - 2
+  );
+  if (remote.finalizedEpoch > maxFinalizedEpoch) {
+    return {
+      code: IrrelevantPeerCode.FINALIZED_EPOCH_TOO_HIGH,
+      finalizedEpoch: remote.finalizedEpoch,
+      maxFinalizedEpoch,
+    };
+  }
+
+  if (remote.finalizedEpoch > GENESIS_EPOCH && isZeroRoot(remote.finalizedRoot)) {
+    return {code: IrrelevantPeerCode.INVALID_FINALIZED_ROOT};
+  }
 
+  // The remote's finalized epoch is equal to ours, but the block root is different to the one in our chain.
+  // Therefore, the node is on a different chain and we should not communicate with them.
   if (
-    remote.finalizedEpoch <= local.finalizedEpoch &&
-    !isZeroRoot(remote.finalizedRoot) &&
-    !isZeroRoot(local.finalizedRoot)
+    remote.finalizedEpoch === local.finalizedEpoch &&
+    !isZeroRoot(local.finalizedRoot) &&
+    !ssz.Root.equals(remote.finalizedRoot, local.finalizedRoot)
   ) {
-    // NOTE: due to preferring to not access chain state here, we can't check the finalized root against our history.
-    // The impact of not doing check is low: peers that are behind us we can't confirm they are in the same chain as us.
-    // In the worst case they will attempt to sync from us, fail and disconnect. The ENR fork check should be sufficient
-    // to differentiate most peers in normal network conditions.
-    const remoteRoot = remote.finalizedRoot;
-    const expectedRoot = remote.finalizedEpoch === local.finalizedEpoch ? local.finalizedRoot : null;
-
-    if (expectedRoot !== null && !ssz.Root.equals(remoteRoot, expectedRoot)) {
-      return {
-        code: IrrelevantPeerCode.DIFFERENT_FINALIZED,
-        expectedRoot: expectedRoot, // forkChoice returns Tree BranchNode which the logger prints as {}
-        remoteRoot: remoteRoot,
-      };
-    }
+    return {
+      code: IrrelevantPeerCode.DIFFERENT_FINALIZED,
+      expectedRoot: local.finalizedRoot,
+      remoteRoot: remote.finalizedRoot,
+    };
   }
 
   if (isForkPostFulu(forkName) && (remote as fulu.Status).earliestAvailableSlot === undefined) {
@@ -76,7 +87,8 @@ export function assertPeerRelevance(
     };
   }
 
-  // Note: Accept request status finalized checkpoint in the future, we do not know if it is a true finalized root
+  // Note: for a behind finalized checkpoint, we validate it more on the main thread against our history,
+  // see BeaconSync.verifyPastFinalizedCheckpoint
   return null;
 }
 
@@ -93,6 +105,10 @@ export function renderIrrelevantPeerType(type: IrrelevantPeerType): string {
       return `DIFFERENT_CLOCKS slotDiff: ${type.slotDiff}`;
     case IrrelevantPeerCode.DIFFERENT_FINALIZED:
       return `DIFFERENT_FINALIZED root: ${toRootHex(type.remoteRoot)} expected: ${toRootHex(type.expectedRoot)}`;
+    case IrrelevantPeerCode.FINALIZED_EPOCH_TOO_HIGH:
+      return `FINALIZED_EPOCH_TOO_HIGH finalizedEpoch: ${type.finalizedEpoch} max: ${type.maxFinalizedEpoch}`;
+    case IrrelevantPeerCode.INVALID_FINALIZED_ROOT:
+      return "INVALID_FINALIZED_ROOT zero root for a non-genesis finalized epoch";
     case IrrelevantPeerCode.NO_EARLIEST_AVAILABLE_SLOT:
       return "No earliestAvailableSlot announced via peer Status";
   }

package/src/network/processor/gossipHandlers.ts

@@ -24,6 +24,7 @@ import {
   deneb,
   fulu,
   gloas,
+  isGloasBeaconBlock,
   isGloasDataColumnSidecar,
   ssz,
   sszTypesFor,
@@ -704,10 +705,8 @@ function getSequentialHandlers(modules: ValidatorFnsModules, options: GossipHand
 
           // Immediately attempt fetch of data columns from execution engine as the bid contains kzg commitments
           // which is all the information we need so there is no reason to delay until execution payload arrives
-          // TODO GLOAS: If we want EL retries after this initial attempt, add an explicit retry policy here
-          // (for example later in the slot). Do not couple retries to incoming gossip columns.
-          // Columns fetched here feed payloadInput.addColumn, which resolves waitForAllData for any
-          // in-flight importExecutionPayload. No processExecutionPayload trigger needed from this path.
+          // Columns fetched here feed payloadInput.addColumn, which resolves the waitForAllData of a
+          // pending payload import. No processExecutionPayload trigger needed from this path.
           chain.getBlobsTracker.triggerGetBlobs(payloadInput);
         }
       })
@@ -717,11 +716,10 @@ function getSequentialHandlers(modules: ValidatorFnsModules, options: GossipHand
 
         if (e instanceof BlockError) {
           switch (e.type.code) {
-            case BlockErrorCode.DATA_UNAVAILABLE: {
-              // Error is quite frequent and not critical
+            case BlockErrorCode.DATA_UNAVAILABLE:
+              // Data did not arrive in time, BlockInputSync keeps pulling it and retries the import
               logLevel = LogLevel.debug;
               break;
-            }
             // ALREADY_KNOWN should not happen with ignoreIfKnown=true above
             // PARENT_BLOCK_UNKNOWN should not happen, we handled this in validateBeaconBlock() function above
             case BlockErrorCode.ALREADY_KNOWN:
@@ -789,7 +787,12 @@ function getSequentialHandlers(modules: ValidatorFnsModules, options: GossipHand
           // roots per proposer and slot, which bounds full imports over gossip to one alternate
           chain.seenBlockProposers.hasBlockRoot(signedBlock.message.slot, e.type.proposerIndex, e.type.root) &&
           // A sibling with an unknown parent cannot be imported, sync fetches it if its branch becomes relevant
-          chain.forkChoice.getBlockHexDefaultStatus(toRootHex(signedBlock.message.parentRoot)) !== null
+          chain.forkChoice.getBlockHexDefaultStatus(toRootHex(signedBlock.message.parentRoot)) !== null &&
+          (!isGloasBeaconBlock(signedBlock.message) ||
+            chain.forkChoice.getBlockHexAndBlockHash(
+              toRootHex(signedBlock.message.parentRoot),
+              toRootHex(signedBlock.message.body.signedExecutionPayloadBid.message.parentBlockHash)
+            ) !== null)
         ) {
           const blockInput = cacheSignatureVerifiedBlock(
             signedBlock,
@@ -924,7 +927,7 @@ function getSequentialHandlers(modules: ValidatorFnsModules, options: GossipHand
             }
 
             // NOTE: we do NOT call chain.processExecutionPayload here. That is triggered only by
-            // envelope arrival (gossip or API). An in-flight importExecutionPayload is awaiting
+            // envelope arrival (gossip or API). A pending payload import is awaiting
             // payloadInput.waitForAllData(); addColumn above will resolve it once hasAllData flips.
 
             scheduleIncompletePayloadEnvelope(payloadInput, peerIdStr);
@@ -1352,6 +1355,10 @@ function getSequentialHandlers(modules: ValidatorFnsModules, options: GossipHand
 
             if (e instanceof PayloadError) {
               switch (e.type.code) {
+                case PayloadErrorCode.DATA_UNAVAILABLE:
+                  // Columns did not arrive in time, BlockInputSync keeps pulling them and retries the import
+                  logLevel = LogLevel.debug;
+                  break;
                 // BLOCK_NOT_IN_FORK_CHOICE should not happen, validateGossipExecutionPayloadEnvelope above
                 // already verified the block is in fork choice
                 case PayloadErrorCode.BLOCK_NOT_IN_FORK_CHOICE:

package/src/node/nodejs.ts

@@ -1,23 +1,29 @@
 import {setMaxListeners} from "node:events";
 import {PrivateKey} from "@libp2p/interface";
 import {Registry} from "prom-client";
+import {hasher} from "@chainsafe/persistent-merkle-tree";
+import {BeaconApiMethods} from "@lodestar/api/beacon/server";
+import {BeaconConfig} from "@lodestar/config";
+import type {LoggerNode} from "@lodestar/logger/node";
 import {
   init as initNativeMetrics,
   registerLocalValidator as registerNativeLocalValidator,
   scrapeMetrics as scrapeNativeMetrics,
   unregisterLocalValidator as unregisterNativeLocalValidator,
-} from "@chainsafe/lodestar-z/metrics";
-import {type PubkeyCache} from "@chainsafe/lodestar-z/pubkeys";
-import {hasher} from "@chainsafe/persistent-merkle-tree";
-import {BeaconApiMethods} from "@lodestar/api/beacon/server";
-import {BeaconConfig} from "@lodestar/config";
-import type {LoggerNode} from "@lodestar/logger/node";
+} from "@lodestar/native/metrics";
+import {type PubkeyCache} from "@lodestar/native/pubkeys";
 import {ZERO_HASH_HEX} from "@lodestar/params";
-import {IBeaconStateView, isStatePostBellatrix, isStatePostGloas} from "@lodestar/state-transition";
+import {
+  IBeaconStateView,
+  computeStartSlotAtEpoch,
+  isStatePostBellatrix,
+  isStatePostGloas,
+} from "@lodestar/state-transition";
 import {Slot} from "@lodestar/types";
 import {sleep, toRootHex} from "@lodestar/utils";
 import {ProcessShutdownCallback} from "@lodestar/validator";
 import {BeaconRestApiServer, getApi} from "../api/index.js";
+import {getDataColumnSidecarsMinEpoch} from "../chain/archiveStore/utils/archiveBlocks.js";
 import {BeaconChain, IBeaconChain, initBeaconMetrics} from "../chain/index.js";
 import {ValidatorMonitor, createValidatorMonitor} from "../chain/validatorMonitor.js";
 import {IBeaconDb} from "../db/index.js";
@@ -212,6 +218,17 @@ export class BeaconNode {
 
     const clock = new Clock({config, genesisTime: anchorState.genesisTime, signal});
 
+    // Before any lookup into the legacy column buckets, which can be slow until compacted
+    const columnsMinEpoch =
+      opts.chain.archiveDataEpochs === Infinity
+        ? null
+        : getDataColumnSidecarsMinEpoch(config, clock.currentEpoch, opts.chain.archiveDataEpochs);
+    await db.compactLegacyDataColumns(
+      columnsMinEpoch !== null && columnsMinEpoch >= config.FULU_FORK_EPOCH
+        ? computeStartSlotAtEpoch(columnsMinEpoch)
+        : 0
+    );
+
     await db.init();
 
     // Prune hot db repos

package/src/node/options.ts

@@ -19,16 +19,15 @@ import {SyncOptions, defaultSyncOptions} from "../sync/options.js";
 
 // Re-export so the CLI doesn't need to depend on lodestar-api
 export {allNamespaces} from "../api/rest/index.js";
-
+export type {EngineApiMode};
 // Re-export to use as default values in CLI args
 export {
-  defaultExecutionEngineHttpOpts,
-  defaultExecutionBuilderHttpOpts,
-  engineApiModes,
   ArchiveMode,
   DEFAULT_ARCHIVE_MODE,
+  defaultExecutionBuilderHttpOpts,
+  defaultExecutionEngineHttpOpts,
+  engineApiModes,
 };
-export type {EngineApiMode};
 
 export interface IBeaconNodeOptions {
   api: ApiOptions;

package/src/sync/interface.ts

@@ -7,6 +7,7 @@ import {IBeaconDb} from "../db/index.js";
 import {Metrics} from "../metrics/index.js";
 import {INetwork} from "../network/index.js";
 import {SyncChainDebugState} from "./range/chain.js";
+
 export type {SyncChainDebugState};
 
 export type SyncingStatus = routes.node.SyncingStatus;

package/src/sync/sync.ts

@@ -1,13 +1,17 @@
 import {SLOTS_PER_EPOCH} from "@lodestar/params";
-import {Slot} from "@lodestar/types";
+import {computeStartSlotAtEpoch} from "@lodestar/state-transition";
+import {Slot, Status} from "@lodestar/types";
 import {Logger, toRootHex} from "@lodestar/utils";
 import {IBeaconChain} from "../chain/index.js";
-import {GENESIS_SLOT} from "../constants/constants.js";
+import {GENESIS_EPOCH, GENESIS_SLOT} from "../constants/constants.js";
+import {GoodByeReasonCode} from "../constants/index.js";
+import {IBeaconDb} from "../db/index.js";
 import {ExecutionEngineState} from "../execution/index.js";
 import {Metrics} from "../metrics/index.js";
-import {INetwork, NetworkEvent, NetworkEventData} from "../network/index.js";
+import {INetwork, NetworkEvent, NetworkEventData, PeerAction} from "../network/index.js";
 import {ClockEvent} from "../util/clock.js";
 import {isOptimisticBlock} from "../util/forkChoice.js";
+import {PeerIdStr} from "../util/peerId.js";
 import {MIN_EPOCH_TO_START_GOSSIP} from "./constants.js";
 import {IBeaconSync, SyncChainDebugState, SyncModules, SyncState, SyncingStatus, syncStateMetric} from "./interface.js";
 import {SyncOptions} from "./options.js";
@@ -19,6 +23,7 @@ export class BeaconSync implements IBeaconSync {
   private readonly logger: Logger;
   private readonly network: INetwork;
   private readonly chain: IBeaconChain;
+  private readonly db: IBeaconDb;
   private readonly metrics: Metrics | null;
   private readonly opts: SyncOptions;
 
@@ -30,10 +35,11 @@ export class BeaconSync implements IBeaconSync {
   private readonly slotImportTolerance: Slot;
 
   constructor(opts: SyncOptions, modules: SyncModules) {
-    const {config, chain, metrics, network, logger} = modules;
+    const {config, chain, db, metrics, network, logger} = modules;
     this.opts = opts;
     this.network = network;
     this.chain = chain;
+    this.db = db;
     this.metrics = metrics;
     this.logger = logger;
     this.rangeSync = new RangeSync(modules, opts);
@@ -192,6 +198,11 @@ export class BeaconSync implements IBeaconSync {
    */
   private addPeer = (data: NetworkEventData[NetworkEvent.peerConnected]): void => {
     const localStatus = this.chain.getStatus();
+    // Genesis and equal finalized epochs are already checked by assertPeerRelevance in the network worker
+    if (data.status.finalizedEpoch > GENESIS_EPOCH && data.status.finalizedEpoch < localStatus.finalizedEpoch) {
+      void this.verifyPastFinalizedCheckpoint(data.peer, data.status);
+    }
+
     const syncType = getPeerSyncType(localStatus, data.status, this.chain.forkChoice, this.slotImportTolerance);
     this.logger.verbose("Peer sync type classified", {
       peer: data.peer,
@@ -216,6 +227,56 @@ export class BeaconSync implements IBeaconSync {
     this.updateSyncState();
   };
 
+  /**
+   * A behind peer's finalized checkpoint must be the canonical block in effect at the start slot of its epoch
+   */
+  async verifyPastFinalizedCheckpoint(peer: PeerIdStr, remote: Status): Promise<void> {
+    const startSlot = computeStartSlotAtEpoch(remote.finalizedEpoch);
+    const root = toRootHex(remote.finalizedRoot);
+    try {
+      let reason: string | null = null;
+      // After a finality jump, finalized blocks stay in fork choice until the archiver migrates them.
+      // Fork choice is pruned only after archiving, so a miss here means the db already has this range.
+      const canonicalBlock = this.chain.forkChoice.getCanonicalBlockClosestLteSlot(startSlot);
+      if (canonicalBlock !== null) {
+        if (canonicalBlock.blockRoot === root) return;
+        reason = "finalized_root_not_canonical";
+      } else {
+        const slot = await this.db.blockArchive.getSlotByRoot(remote.finalizedRoot);
+        if (slot !== null && slot > startSlot) {
+          reason = "finalized_root_after_epoch_start";
+        } else if (slot !== startSlot) {
+          const childSlot = await this.db.blockArchive.getSlotByParentRoot(remote.finalizedRoot);
+          if (childSlot !== null) {
+            if (childSlot <= startSlot) reason = "finalized_root_not_in_effect_at_epoch_start";
+          } else if (slot === null && startSlot > this.chain.earliestAvailableSlot) {
+            // childSlot == null too, neither the root nor a child is canonical
+            // or they are canonical and both pruned but peer's finalized epoch starts after earliestAvailableSlot
+            reason = "unknown_finalized_root";
+          }
+        }
+      }
+
+      if (reason !== null) {
+        this.logger.debug("Peer finalized checkpoint not in our chain", {
+          peer,
+          finalizedEpoch: remote.finalizedEpoch,
+          root,
+          reason,
+        });
+        this.metrics?.syncPeerFinalizedCheckpointMismatch.inc({reason});
+        this.network.reportPeer(peer, PeerAction.Fatal, "PeerFinalizedCheckpointMismatch");
+        await this.network.goodbyeAndDisconnect(peer, GoodByeReasonCode.IRRELEVANT_NETWORK);
+      }
+    } catch (e) {
+      this.logger.debug(
+        "Error verifying peer finalized checkpoint",
+        {peer, finalizedEpoch: remote.finalizedEpoch, root},
+        e as Error
+      );
+    }
+  }
+
   /**
    * Must be called by libp2p when a peer is removed from the peer manager
    */

package/src/util/file.ts

@@ -1,25 +1,24 @@
-import fs from "node:fs";
-import {promisify} from "node:util";
+import {readFile as fsReadFile, mkdir, readdir, stat, unlink, writeFile} from "node:fs/promises";
 
 /** Ensure a directory exists */
 export async function ensureDir(path: string): Promise<void> {
   try {
-    await promisify(fs.stat)(path);
+    await stat(path);
   } catch (_) {
     // not exists
-    await promisify(fs.mkdir)(path, {recursive: true});
+    await mkdir(path, {recursive: true});
   }
 }
 
 /** Write data to a file if it does not exist */
 export async function writeIfNotExist(filepath: string, bytes: Uint8Array): Promise<boolean> {
   try {
-    await promisify(fs.stat)(filepath);
+    await stat(filepath);
     return false;
     // file exists, do nothing
   } catch (_) {
     // not exists
-    await promisify(fs.writeFile)(filepath, bytes);
+    await writeFile(filepath, bytes);
     return true;
   }
 }
@@ -27,7 +26,7 @@ export async function writeIfNotExist(filepath: string, bytes: Uint8Array): Prom
 /** Remove a file if it exists */
 export async function removeFile(path: string): Promise<boolean> {
   try {
-    await promisify(fs.unlink)(path);
+    await unlink(path);
     return true;
   } catch (_) {
     // may not exists
@@ -37,7 +36,7 @@ export async function removeFile(path: string): Promise<boolean> {
 
 export async function readFile(path: string): Promise<Uint8Array | null> {
   try {
-    return await fs.promises.readFile(path);
+    return await fsReadFile(path);
   } catch (_) {
     return null;
   }
@@ -45,7 +44,7 @@ export async function readFile(path: string): Promise<Uint8Array | null> {
 
 export async function readFileNames(folderPath: string): Promise<string[]> {
   try {
-    return await fs.promises.readdir(folderPath);
+    return await readdir(folderPath);
   } catch (_) {
     return [];
   }

package/src/util/queue/itemQueue.ts

@@ -69,7 +69,7 @@ export class JobItemQueue<Args extends any[], R> {
       if (this.jobs.length === 1 && this.opts.noYieldIfOneItem) {
         void this.runJob();
       } else if (this.runningJobs < this.opts.maxConcurrency) {
-        callInNextEventLoop(this.runJob);
+        callInNextEventLoop(() => void this.runJob());
       }
     });
   }

package/src/util/sszBytes.ts

@@ -854,7 +854,7 @@ export function getBlobKzgCommitmentsCountFromSignedBeaconBlockSerialized(
 
     const start = bidMsgStart + commitmentsRange.start;
     const end = bidMsgStart + commitmentsRange.end;
-    return Math.round(((end > blockBytes.byteLength ? blockBytes.byteLength : end) - start) / commitmentSize);
+    return Math.round((Math.min(end, blockBytes.byteLength) - start) / commitmentSize);
   }
 
   const {SignedBeaconBlock, BeaconBlock, BeaconBlockBody, KZGCommitment} = ssz[forkName as ForkPostDeneb];
@@ -880,5 +880,5 @@ export function getBlobKzgCommitmentsCountFromSignedBeaconBlockSerialized(
   const end = messageRange.end + bodyRange.end + kzgCommitmentsRange.end;
   const start = messageRange.start + bodyRange.start + kzgCommitmentsRange.start;
 
-  return Math.round(((end > blockBytes.byteLength ? blockBytes.byteLength : end) - start) / commitmentSize);
+  return Math.round((Math.min(end, blockBytes.byteLength) - start) / commitmentSize);
 }