@lodestar/beacon-node 1.47.0-dev.7481aa5850 → 1.47.0-dev.7aa8c9c93a

package/src/metrics/metrics/lodestar.ts

@@ -4,7 +4,10 @@ import {ArchiveStoreTask} from "../../chain/archiveStore/archiveStore.js";
 import {FrequencyStateArchiveStep} from "../../chain/archiveStore/strategies/frequencyStateArchiveStrategy.js";
 import {BlockInputSource} from "../../chain/blocks/blockInput/index.js";
 import {PayloadErrorCode} from "../../chain/blocks/importExecutionPayload.js";
-import {PayloadEnvelopeInputSource} from "../../chain/blocks/payloadEnvelopeInput/index.js";
+import {
+  PayloadEnvelopeInputPruneReason,
+  PayloadEnvelopeInputSource,
+} from "../../chain/blocks/payloadEnvelopeInput/index.js";
 import {JobQueueItemType} from "../../chain/bls/index.js";
 import {AttestationErrorCode, BlockErrorCode} from "../../chain/errors/index.js";
 import {
@@ -31,6 +34,10 @@ import {RegistryMetricCreator} from "../utils/registryMetricCreator.js";
 
 export type LodestarMetrics = ReturnType<typeof createLodestarMetrics>;
 
+type BlsJobOutcome = "valid" | "invalid" | "prepError" | "verifyError" | "workerError";
+type BlsBufferFlushReason = "size" | "timeout";
+type BlsVerificationCallOperation = "batch" | "single" | "fallback" | "same_message";
+
 /**
  * Extra Lodestar custom metrics
  */
@@ -396,86 +403,108 @@ export function createLodestarMetrics(
     // BLS verifier thread pool and queue
 
     bls: {
-      aggregatedPubkeys: register.gauge({
+      aggregatedPubkeys: register.counter({
         name: "lodestar_bls_aggregated_pubkeys_total",
-        help: "Total aggregated pubkeys for BLS validation",
+        help: "Total validator public keys used as inputs to aggregate signature sets",
+      }),
+      pubkeyCacheSize: register.gauge({
+        name: "lodestar_bls_pubkey_cache_size",
+        help: "Current number of validator public keys in the native cache",
+      }),
+      pubkeyCacheCapacity: register.gauge({
+        name: "lodestar_bls_pubkey_cache_capacity",
+        help: "Number of validator public keys the native cache can hold without growing",
       }),
     },
 
     blsThreadPool: {
-      jobsWorkerTime: register.gauge<{workerId: number}>({
+      jobsWorkerTime: register.counter<{workerId: number}>({
         name: "lodestar_bls_thread_pool_time_seconds_sum",
-        help: "Total time spent verifying signature sets measured on the worker",
+        help: "Cumulative wall time spent processing BLS dispatch groups by worker",
         labelNames: ["workerId"],
       }),
-      successJobsSignatureSetsCount: register.gauge({
+      successJobsSignatureSetsCount: register.counter({
         name: "lodestar_bls_thread_pool_success_jobs_signature_sets_count",
-        help: "Count of total verified signature sets",
-      }),
-      errorAggregateSignatureSetsCount: register.gauge<{type: JobQueueItemType}>({
-        name: "lodestar_bls_thread_pool_error_aggregate_signature_sets_count",
-        help: "Count of error when aggregating pubkeys or signatures",
-        labelNames: ["type"],
+        help: "Count of signature sets completed without an operational error",
       }),
-      errorJobsSignatureSetsCount: register.gauge({
+      errorJobsSignatureSetsCount: register.counter({
         name: "lodestar_bls_thread_pool_error_jobs_signature_sets_count",
-        help: "Count of total error-ed signature sets",
+        help: "Count of signature sets that ended in an input, binding, or worker error",
       }),
-      jobWaitTime: register.histogram({
+      jobWaitTime: register.histogram<{type: JobQueueItemType}>({
         name: "lodestar_bls_thread_pool_queue_job_wait_time_seconds",
-        help: "Time from job added to the queue to starting the job in seconds",
-        buckets: [0.01, 0.02, 0.1, 0.3, 0.5, 1],
+        help: "Time from adding a BLS job until it is selected for a worker dispatch group",
+        labelNames: ["type"],
+        buckets: [0.005, 0.01, 0.025, 0.05, 0.1, 0.2, 0.5, 1, 2],
+      }),
+      jobDuration: register.histogram<{type: JobQueueItemType}>({
+        name: "lodestar_bls_thread_pool_job_duration_seconds",
+        help: "End-to-end duration of a BLS job from submission through completion",
+        labelNames: ["type"],
+        buckets: [0.005, 0.01, 0.025, 0.05, 0.1, 0.2, 0.5, 1, 2],
+      }),
+      jobResults: register.counter<{type: JobQueueItemType; outcome: BlsJobOutcome}>({
+        name: "lodestar_bls_thread_pool_job_results_total",
+        help: "Count of completed BLS jobs by cryptographic or operational outcome",
+        labelNames: ["type", "outcome"],
       }),
       queueLength: register.gauge({
         name: "lodestar_bls_thread_pool_queue_length",
-        help: "Count of total block processor queue length",
+        help: "Count of ready BLS jobs waiting for a worker",
+      }),
+      bufferedQueueLength: register.gauge({
+        name: "lodestar_bls_thread_pool_buffered_queue_length",
+        help: "Count of BLS jobs waiting in the intentional batching buffer",
       }),
       workersBusy: register.gauge({
         name: "lodestar_bls_thread_pool_workers_busy",
         help: "Count of current busy workers",
       }),
-      totalJobsGroupsStarted: register.gauge({
+      totalJobsGroupsStarted: register.counter({
         name: "lodestar_bls_thread_pool_job_groups_started_total",
-        help: "Count of total jobs groups started in bls thread pool, job groups include +1 jobs",
+        help: "Count of worker dispatch groups started by the BLS thread pool",
       }),
-      totalJobsStarted: register.gauge<{type: JobQueueItemType}>({
+      totalJobsStarted: register.counter<{type: JobQueueItemType}>({
         name: "lodestar_bls_thread_pool_jobs_started_total",
-        help: "Count of total jobs started in bls thread pool, jobs include +1 signature sets",
+        help: "Count of jobs started by the BLS thread pool",
         labelNames: ["type"],
       }),
-      totalSigSetsStarted: register.gauge<{type: JobQueueItemType}>({
+      totalSigSetsStarted: register.counter<{type: JobQueueItemType}>({
         name: "lodestar_bls_thread_pool_sig_sets_started_total",
-        help: "Count of total signature sets started in bls thread pool, sig sets include 1 pk, msg, sig",
+        help: "Count of signature sets started by the BLS thread pool",
         labelNames: ["type"],
       }),
       // Re-verifying a batch means doing double work. This number must be very low or it can be a waste of CPU resources
-      batchRetries: register.gauge({
+      batchRetries: register.counter({
         name: "lodestar_bls_thread_pool_batch_retries_total",
-        help: "Count of total batches that failed and had to be verified again.",
+        help: "Count of combined batches that failed and required per-job fallback",
       }),
       // To count how many sigs are being validated with the optimization of batching them
-      batchSigsSuccess: register.gauge({
+      batchSigsSuccess: register.counter({
         name: "lodestar_bls_thread_pool_batch_sigs_success_total",
-        help: "Count of total batches that failed and had to be verified again.",
+        help: "Count of signature sets accepted by combined batch verification without fallback",
       }),
-      sameMessageRetryJobs: register.gauge({
-        name: "lodestar_bls_thread_pool_same_message_jobs_retries_total",
-        help: "Count of total same message jobs that failed and had to be verified again.",
+      bufferFlushes: register.counter<{reason: BlsBufferFlushReason}>({
+        name: "lodestar_bls_thread_pool_buffer_flushes_total",
+        help: "Count of batching-buffer flushes by trigger",
+        labelNames: ["reason"],
       }),
-      sameMessageRetrySets: register.gauge({
-        name: "lodestar_bls_thread_pool_same_message_sets_retries_total",
-        help: "Count of total same message sets that failed and had to be verified again.",
+      bufferedSignatureSets: register.histogram({
+        name: "lodestar_bls_thread_pool_buffered_signature_sets",
+        help: "Count of signature sets included in each batching-buffer flush",
+        buckets: [1, 8, 16, 24, 28, 32, 36, 40, 44, 48, 64, 128],
       }),
       // To measure the time cost of main thread <-> worker message passing
       latencyToWorker: register.histogram({
         name: "lodestar_bls_thread_pool_latency_to_worker",
         help: "Time from sending the job to the worker and the worker receiving it",
-        buckets: [0.001, 0.003, 0.01, 0.03, 0.1],
+        // The dashboard treats 20ms as slow; values above 100ms only need severe-stall classification.
+        buckets: [0.00025, 0.001, 0.005, 0.02, 0.1],
       }),
       latencyFromWorker: register.histogram({
         name: "lodestar_bls_thread_pool_latency_from_worker",
         help: "Time from the worker sending the result and the main thread receiving it",
-        buckets: [0.001, 0.003, 0.01, 0.03, 0.1],
+        buckets: [0.00025, 0.001, 0.005, 0.02, 0.1],
       }),
       mainThreadDurationInThreadPool: register.histogram({
         name: "lodestar_bls_thread_pool_main_thread_time_seconds",
@@ -485,32 +514,40 @@ export function createLodestarMetrics(
       }),
       timePerSigSet: register.histogram({
         name: "lodestar_bls_worker_thread_time_per_sigset_seconds",
-        help: "Time to verify each sigset with worker thread mode",
+        help: "Average BLS worker dispatch wall time per original signature set",
         // Time per sig ~0.9ms on good machines
-        buckets: [0.5e-3, 0.75e-3, 1e-3, 1.5e-3, 2e-3, 5e-3],
-      }),
-      totalSigSets: register.gauge({
+        buckets: [0.5e-3, 0.75e-3, 1e-3, 1.5e-3, 2e-3, 5e-3, 10e-3, 20e-3, 50e-3, 100e-3],
+      }),
+      workRequestPreparationDuration: register.histogram({
+        name: "lodestar_bls_thread_pool_work_request_preparation_duration_seconds",
+        help: "Main-thread time spent converting one worker dispatch group to binding inputs",
+        // Normal requests take microseconds; the dashboard treats 1ms as a main-thread stall.
+        buckets: [0.000005, 0.000025, 0.0001, 0.0005, 0.001, 0.005, 0.025],
+      }),
+      verificationCallDuration: register.histogram<{operation: BlsVerificationCallOperation}>({
+        name: "lodestar_bls_thread_pool_verification_call_duration_seconds",
+        help: "Duration of one worker-side lodestar-z verification binding call, including input conversion",
+        labelNames: ["operation"],
+        // Verification ranges from sub-millisecond calls to large batches; above 500ms is uniformly severe.
+        buckets: [0.00025, 0.001, 0.005, 0.02, 0.1, 0.5],
+      }),
+      verificationCallSignatureSets: register.counter<{operation: BlsVerificationCallOperation}>({
+        name: "lodestar_bls_thread_pool_verification_call_signature_sets_total",
+        help: "Count of signature sets passed to worker-side lodestar-z verification binding calls",
+        labelNames: ["operation"],
+      }),
+      totalSigSets: register.counter({
         name: "lodestar_bls_thread_pool_sig_sets_total",
         help: "Count of total signature sets",
       }),
-      prioritizedSigSets: register.gauge({
+      prioritizedSigSets: register.counter({
         name: "lodestar_bls_thread_pool_prioritized_sig_sets_total",
         help: "Count of total prioritized signature sets",
       }),
-      batchableSigSets: register.gauge({
+      batchableSigSets: register.counter({
         name: "lodestar_bls_thread_pool_batchable_sig_sets_total",
         help: "Count of total batchable signature sets",
       }),
-      aggregateWithRandomnessAsyncDuration: register.histogram({
-        name: "lodestar_bls_thread_pool_aggregate_with_randomness_async_time_seconds",
-        help: "Total time performing aggregateWithRandomness async",
-        buckets: [0.001, 0.005, 0.01, 0.1, 0.3],
-      }),
-      pubkeysAggregationMainThreadDuration: register.histogram({
-        name: "lodestar_bls_thread_pool_pubkeys_aggregation_main_thread_time_seconds",
-        help: "Total time spent aggregating pubkeys on main thread",
-        buckets: [0.001, 0.005, 0.01, 0.1],
-      }),
     },
 
     // BLS time on single thread mode
@@ -1609,9 +1646,15 @@ export function createLodestarMetrics(
           name: "lodestar_seen_payload_envelope_input_cache_serialized_object_refs",
           help: "Number of serialized-cache object refs retained by cached PayloadEnvelopeInputs",
         }),
-        created: register.counter({
+        created: register.counter<{source: PayloadEnvelopeInputSource}>({
           name: "lodestar_seen_payload_envelope_input_cache_items_created_total",
-          help: "Number of PayloadEnvelopeInputs created",
+          help: "Number of PayloadEnvelopeInputs created by source",
+          labelNames: ["source"],
+        }),
+        pruned: register.counter<{reason: PayloadEnvelopeInputPruneReason}>({
+          name: "lodestar_seen_payload_envelope_input_cache_items_pruned_total",
+          help: "Number of PayloadEnvelopeInputs evicted by reason",
+          labelNames: ["reason"],
         }),
       },
     },

package/src/network/gossip/topic.ts

@@ -414,3 +414,41 @@ export const gossipTopicIgnoreDuplicatePublishError: Record<GossipType, boolean>
   [GossipType.execution_payload_bid]: true,
   [GossipType.proposer_preferences]: true,
 };
+
+/**
+ * Whether a publish that reached zero subscribed peers is an acceptable outcome for a topic.
+ *
+ * Publishing to a topic with no subscribed peers throws `PublishError.NoPeersSubscribedToTopic`, which
+ * propagates to the caller and, for messages submitted through the beacon API, is surfaced to the
+ * validator client. That is the right default: a message that reached nobody is a failed broadcast, and a
+ * VC configured with several beacon nodes can fall back to another one.
+ *
+ * Only topics that have a concrete reason to tolerate it opt out here, ie. are set to `true`, and the
+ * reason is named on the entry. Everything else is `false` and keeps raising. Note this is applied per
+ * publish, a topic set to `false` still honors the global `--network.allowPublishToZeroPeers` flag.
+ */
+export const gossipTopicAllowPublishToZeroPeers: Record<GossipType, boolean> = {
+  [GossipType.beacon_block]: false,
+  [GossipType.blob_sidecar]: false,
+  // data_column_sidecar: we ensure having all topic peers via prioritizePeers(). Even with 0 peers on the
+  // topic the overall publish can still succeed, because a supernode rebuilds and publishes the missing
+  // columns for us, so track sent peers as 0 instead of raising
+  [GossipType.data_column_sidecar]: true,
+  [GossipType.beacon_aggregate_and_proof]: false,
+  [GossipType.beacon_attestation]: false,
+  [GossipType.voluntary_exit]: false,
+  [GossipType.proposer_slashing]: false,
+  [GossipType.attester_slashing]: false,
+  [GossipType.sync_committee_contribution_and_proof]: false,
+  [GossipType.sync_committee]: false,
+  // light_client_finality_update and light_client_optimistic_update: non-mandatory route on most of the
+  // network as of Oct 2022, there may be no peer subscribed to the topic at all. Reason carried over from
+  // the handlers in network.ts, which used to swallow the error instead
+  [GossipType.light_client_finality_update]: true,
+  [GossipType.light_client_optimistic_update]: true,
+  [GossipType.bls_to_execution_change]: false,
+  [GossipType.execution_payload]: false,
+  [GossipType.payload_attestation_message]: false,
+  [GossipType.execution_payload_bid]: false,
+  [GossipType.proposer_preferences]: false,
+};

package/src/network/libp2p/index.ts

@@ -1,3 +1,4 @@
+import {getCiphers} from "node:crypto";
 import {bootstrap} from "@libp2p/bootstrap";
 import {identify} from "@libp2p/identify";
 import type {PrivateKey} from "@libp2p/interface";
@@ -9,12 +10,20 @@ import {Libp2pInit, createLibp2p} from "libp2p";
 import {Registry} from "prom-client";
 import {ENR} from "@chainsafe/enr";
 import {noise} from "@chainsafe/libp2p-noise";
-import {defaultCrypto} from "@chainsafe/libp2p-noise/crypto";
+import {asCrypto, defaultCrypto} from "@chainsafe/libp2p-noise/crypto";
 import {quic} from "@chainsafe/libp2p-quic";
 import {Libp2p, LodestarComponents} from "../interface.js";
 import {NetworkOptions, defaultNetworkOptions} from "../options.js";
 import {Eth2PeerDataStore} from "../peers/datastore.js";
 
+const noiseCrypto = getCiphers().includes("chacha20-poly1305")
+  ? defaultCrypto
+  : {
+      ...defaultCrypto,
+      chaCha20Poly1305Encrypt: asCrypto.chaCha20Poly1305Encrypt,
+      chaCha20Poly1305Decrypt: asCrypto.chaCha20Poly1305Decrypt,
+    };
+
 export type NodeJsLibp2pOpts = {
   peerStoreDir?: string;
   disablePeerDiscovery?: boolean;
@@ -122,7 +131,7 @@ export async function createNodeJsLibp2p(
       listen: localMultiaddrs,
       announce: [],
     },
-    connectionEncrypters: [noise({crypto: defaultCrypto})],
+    connectionEncrypters: [noise({crypto: noiseCrypto})],
     transports,
     streamMuxers: [mplex({disconnectThreshold})],
     peerDiscovery,

package/src/network/network.ts

@@ -49,7 +49,12 @@ import {INetworkCore, NetworkCore, WorkerNetworkCore} from "./core/index.js";
 import {INetworkEventBus, NetworkEvent, NetworkEventBus, NetworkEventData} from "./events.js";
 import {getActiveForkBoundaries} from "./forks.js";
 import {GossipHandlers, GossipTopicMap, GossipType, GossipTypeMap} from "./gossip/index.js";
-import {getGossipSSZType, gossipTopicIgnoreDuplicatePublishError, stringifyGossipTopic} from "./gossip/topic.js";
+import {
+  getGossipSSZType,
+  gossipTopicAllowPublishToZeroPeers,
+  gossipTopicIgnoreDuplicatePublishError,
+  stringifyGossipTopic,
+} from "./gossip/topic.js";
 import {INetwork} from "./interface.js";
 import {NetworkOptions} from "./options.js";
 import {PeerAction, PeerScoreStats} from "./peers/index.js";
@@ -65,7 +70,7 @@ import {
 } from "./reqresp/utils/collect.js";
 import {collectSequentialBlocksInRange} from "./reqresp/utils/collectSequentialBlocksInRange.js";
 import {CommitteeSubscription} from "./subnets/index.js";
-import {isPublishDuplicateError, isPublishToZeroPeersError, prettyPrintPeerIdStr} from "./util.js";
+import {isPublishDuplicateError, prettyPrintPeerIdStr} from "./util.js";
 
 type NetworkModules = {
   opts: NetworkOptions;
@@ -352,9 +357,7 @@ export class Network implements INetwork {
     const epoch = computeEpochAtSlot(signedBlock.message.slot);
     const boundary = this.config.getForkBoundaryAtEpoch(epoch);
 
-    return this.publishGossip<GossipType.beacon_block>({type: GossipType.beacon_block, boundary}, signedBlock, {
-      ignoreDuplicatePublishError: true,
-    });
+    return this.publishGossip<GossipType.beacon_block>({type: GossipType.beacon_block, boundary}, signedBlock);
   }
 
   async publishBlobSidecar(blobSidecar: deneb.BlobSidecar): Promise<number> {
@@ -363,9 +366,7 @@ export class Network implements INetwork {
 
     const subnet = blobSidecar.index;
 
-    return this.publishGossip<GossipType.blob_sidecar>({type: GossipType.blob_sidecar, boundary, subnet}, blobSidecar, {
-      ignoreDuplicatePublishError: true,
-    });
+    return this.publishGossip<GossipType.blob_sidecar>({type: GossipType.blob_sidecar, boundary, subnet}, blobSidecar);
   }
 
   async publishDataColumnSidecar(
@@ -381,14 +382,7 @@ export class Network implements INetwork {
     try {
       const sentPeers = await this.publishGossip<GossipType.data_column_sidecar>(
         {type: GossipType.data_column_sidecar, boundary, subnet},
-        dataColumnSidecar,
-        {
-          // we ensure having all topic peers via prioritizePeers() function
-          // in the worse case, if there is 0 peer on the topic, the overall publish operation could be still a success
-          // because supernode will rebuild and publish missing data column sidecars for us
-          // hence we want to track sent peers as 0 instead of an error
-          allowPublishToZeroTopicPeers: true,
-        }
+        dataColumnSidecar
       );
       return {sentPeers, alreadyPublished: false};
     } catch (e) {
@@ -405,8 +399,7 @@ export class Network implements INetwork {
 
     return this.publishGossip<GossipType.beacon_aggregate_and_proof>(
       {type: GossipType.beacon_aggregate_and_proof, boundary},
-      aggregateAndProof,
-      {ignoreDuplicatePublishError: true}
+      aggregateAndProof
     );
   }
 
@@ -416,8 +409,7 @@ export class Network implements INetwork {
 
     return this.publishGossip<GossipType.beacon_attestation>(
       {type: GossipType.beacon_attestation, boundary, subnet},
-      attestation,
-      {ignoreDuplicatePublishError: true}
+      attestation
     );
   }
 
@@ -425,9 +417,7 @@ export class Network implements INetwork {
     const epoch = voluntaryExit.message.epoch;
     const boundary = this.config.getForkBoundaryAtEpoch(epoch);
 
-    return this.publishGossip<GossipType.voluntary_exit>({type: GossipType.voluntary_exit, boundary}, voluntaryExit, {
-      ignoreDuplicatePublishError: true,
-    });
+    return this.publishGossip<GossipType.voluntary_exit>({type: GossipType.voluntary_exit, boundary}, voluntaryExit);
   }
 
   async publishBlsToExecutionChange(blsToExecutionChange: capella.SignedBLSToExecutionChange): Promise<number> {
@@ -438,8 +428,7 @@ export class Network implements INetwork {
       if (fork >= ForkSeq.capella) {
         const publishPromise = this.publishGossip<GossipType.bls_to_execution_change>(
           {type: GossipType.bls_to_execution_change, boundary},
-          blsToExecutionChange,
-          {ignoreDuplicatePublishError: true}
+          blsToExecutionChange
         );
         publishChanges.push(publishPromise);
       }
@@ -457,8 +446,7 @@ export class Network implements INetwork {
 
     return this.publishGossip<GossipType.proposer_slashing>(
       {type: GossipType.proposer_slashing, boundary},
-      proposerSlashing,
-      {ignoreDuplicatePublishError: true}
+      proposerSlashing
     );
   }
 
@@ -478,10 +466,7 @@ export class Network implements INetwork {
 
     return this.publishGossip<GossipType.sync_committee>(
       {type: GossipType.sync_committee, boundary, subnet},
-      signature,
-      {
-        ignoreDuplicatePublishError: true,
-      }
+      signature
     );
   }
 
@@ -491,8 +476,7 @@ export class Network implements INetwork {
 
     return this.publishGossip<GossipType.sync_committee_contribution_and_proof>(
       {type: GossipType.sync_committee_contribution_and_proof, boundary},
-      contributionAndProof,
-      {ignoreDuplicatePublishError: true}
+      contributionAndProof
     );
   }
 
@@ -522,8 +506,7 @@ export class Network implements INetwork {
 
     return this.publishGossip<GossipType.execution_payload>(
       {type: GossipType.execution_payload, boundary},
-      signedEnvelope,
-      {ignoreDuplicatePublishError: true}
+      signedEnvelope
     );
   }
 
@@ -533,8 +516,7 @@ export class Network implements INetwork {
 
     return this.publishGossip<GossipType.execution_payload_bid>(
       {type: GossipType.execution_payload_bid, boundary},
-      signedBid,
-      {ignoreDuplicatePublishError: true}
+      signedBid
     );
   }
 
@@ -544,8 +526,7 @@ export class Network implements INetwork {
 
     return this.publishGossip<GossipType.payload_attestation_message>(
       {type: GossipType.payload_attestation_message, boundary},
-      payloadAttestationMessage,
-      {ignoreDuplicatePublishError: true}
+      payloadAttestationMessage
     );
   }
 
@@ -555,22 +536,21 @@ export class Network implements INetwork {
 
     return this.publishGossip<GossipType.proposer_preferences>(
       {type: GossipType.proposer_preferences, boundary},
-      signedProposerPreferences,
-      {ignoreDuplicatePublishError: true}
+      signedProposerPreferences
     );
   }
 
   private async publishGossip<K extends GossipType>(
     topic: GossipTopicMap[K],
-    object: GossipTypeMap[K],
-    opts?: PublishOpts | undefined
+    object: GossipTypeMap[K]
   ): Promise<number> {
     const topicStr = stringifyGossipTopic(this.config, topic);
     const sszType = getGossipSSZType(topic);
     const messageData = (sszType.serialize as (object: GossipTypeMap[GossipType]) => Uint8Array)(object);
-    opts = {
-      ...opts,
+    const opts: PublishOpts = {
       ignoreDuplicatePublishError: gossipTopicIgnoreDuplicatePublishError[topic.type],
+      // Leave undefined unless the topic opts out, so `--network.allowPublishToZeroPeers` still applies
+      allowPublishToZeroTopicPeers: gossipTopicAllowPublishToZeroPeers[topic.type] ? true : undefined,
     };
     const sentPeers = await this.core.publishGossip(topicStr, messageData, opts);
 
@@ -813,11 +793,7 @@ export class Network implements INetwork {
       await this.waitForSyncMessageCutoff(finalityUpdate.signatureSlot);
       await this.publishLightClientFinalityUpdate(finalityUpdate);
     } catch (e) {
-      // Non-mandatory route on most of network as of Oct 2022. May not have found any peers on topic yet
-      // Remove once https://github.com/ChainSafe/js-libp2p-gossipsub/issues/367
-      if (!isPublishToZeroPeersError(e as Error)) {
-        this.logger.debug("Error on BeaconGossipHandler.onLightclientFinalityUpdate", {}, e as Error);
-      }
+      this.logger.debug("Error on BeaconGossipHandler.onLightclientFinalityUpdate", {}, e as Error);
     }
   };
 
@@ -830,11 +806,7 @@ export class Network implements INetwork {
       await this.waitForSyncMessageCutoff(optimisticUpdate.signatureSlot);
       await this.publishLightClientOptimisticUpdate(optimisticUpdate);
     } catch (e) {
-      // Non-mandatory route on most of network as of Oct 2022. May not have found any peers on topic yet
-      // Remove once https://github.com/ChainSafe/js-libp2p-gossipsub/issues/367
-      if (!isPublishToZeroPeersError(e as Error)) {
-        this.logger.debug("Error on BeaconGossipHandler.onLightclientOptimisticUpdate", {}, e as Error);
-      }
+      this.logger.debug("Error on BeaconGossipHandler.onLightclientOptimisticUpdate", {}, e as Error);
     }
   };
 

package/src/network/processor/gossipHandlers.ts

@@ -196,7 +196,8 @@ function getSequentialHandlers(modules: ValidatorFnsModules, options: GossipHand
         forkName: fork,
         sampledColumns: chain.custodyConfig.sampledColumns,
         custodyColumns: chain.custodyConfig.custodyColumns,
-        timeCreatedSec: seenTimestampSec,
+        seenTimestampSec,
+        source: PayloadEnvelopeInputSource.gossip,
       });
     }
 
@@ -640,6 +641,7 @@ function getSequentialHandlers(modules: ValidatorFnsModules, options: GossipHand
         metrics?.gossipBlock.elapsedTimeTillProcessed.observe(delaySec);
 
         if (isForkPostGloas(blockInput.forkName)) {
+          // we should have the payloadInput in the seen cache so no need getOrReload() here
           const payloadInput = chain.seenPayloadEnvelopeInputCache.get(blockInput.blockRootHex);
           // This payloadInput should have been created just after gossip validation
           if (!payloadInput) {
@@ -1189,6 +1191,8 @@ function getSequentialHandlers(modules: ValidatorFnsModules, options: GossipHand
       chain.validatorMonitor?.registerExecutionPayloadEnvelope(OpSource.gossip, delaySec, signedEnvelope);
 
       const blockRootHex = toRootHex(envelope.beaconBlockRoot);
+      // a gossip payload cannot be `MAX_PAYLOAD_ENVELOPE_INPUT_CACHE_SIZE` slots after the block
+      // otherwise it'll get to UnknownBlockInput flow
       const payloadInput = chain.seenPayloadEnvelopeInputCache.get(blockRootHex);
 
       if (!payloadInput) {

package/src/network/reqresp/score.ts

@@ -44,17 +44,11 @@ export function onOutgoingReqRespError(e: RequestError, method: ReqRespMethod):
         // failure, so keep the stronger penalty (Fatal for Ping, as before).
         return method === ReqRespMethod.Ping ? PeerAction.Fatal : PeerAction.LowToleranceError;
       }
-      switch (method) {
-        // Ping and Status are liveness probes; their dial timeouts are dominated by transient
-        // network congestion rather than peer misbehavior, so penalize leniently to avoid
-        // self-inflicted peer starvation (https://github.com/ChainSafe/lodestar/issues/9562),
-        // while still applying some penalty so genuinely dead peers eventually free the slot.
-        case ReqRespMethod.Ping:
-        case ReqRespMethod.Status:
-          return PeerAction.HighToleranceError;
-        default:
-          return PeerAction.LowToleranceError;
-      }
+      // A dial failure means we could not open a stream. That is evidence about the transport, not
+      // about the peer's behavior, and it is dominated by transient congestion on our side, so
+      // penalize leniently for every method to avoid self-inflicted peer starvation (#9562). A
+      // genuinely dead peer still accumulates penalty and is eventually evicted, freeing the slot.
+      return PeerAction.HighToleranceError;
     // TODO: Detect SSZDecodeError and return PeerAction.Fatal
 
     case RequestErrorCode.RESP_TIMEOUT:

package/src/network/util.ts

@@ -23,11 +23,6 @@ export function getConnection(libp2p: Libp2p, peerIdStr: string): Connection | u
   return getConnectionsMap(libp2p).get(peerIdStr)?.value[0] ?? undefined;
 }
 
-// https://github.com/libp2p/js-libp2p/blob/f87cba928991736d9646b3e054c367f55cab315c/packages/gossipsub/src/gossipsub.ts#L2076
-export function isPublishToZeroPeersError(e: Error): boolean {
-  return e.message.includes("PublishError.NoPeersSubscribedToTopic");
-}
-
 // https://github.com/libp2p/js-libp2p/blob/f87cba928991736d9646b3e054c367f55cab315c/packages/gossipsub/src/gossipsub.ts#L2076
 export function isPublishDuplicateError(e: Error): boolean {
   return e.message.includes("PublishError.Duplicate");

package/src/node/nodejs.ts

@@ -1,12 +1,13 @@
 import {setMaxListeners} from "node:events";
 import {PrivateKey} from "@libp2p/interface";
 import {Registry} from "prom-client";
+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";
 import {ZERO_HASH_HEX} from "@lodestar/params";
-import {IBeaconStateView, PubkeyCache, isStatePostBellatrix, isStatePostGloas} from "@lodestar/state-transition";
+import {IBeaconStateView, isStatePostBellatrix, isStatePostGloas} from "@lodestar/state-transition";
 import {phase0} from "@lodestar/types";
 import {sleep, toRootHex} from "@lodestar/utils";
 import {ProcessShutdownCallback} from "@lodestar/validator";