timeout increase

Author: sekuba

collector/.env.example

@@ -21,6 +21,8 @@ L1_RPC_URL=http://127.0.0.1:8545
 # L1_SLASH_LOG_START_BLOCK=25533241
 # Used only when no exact start block is configured.
 L1_SLASH_LOG_LOOKBACK_BLOCKS=50000
+# Archive calls for a chunk containing slash executions can take longer.
+L1_SLASH_LOG_PROVIDER_TIMEOUT_MS=30000
 
 # Enable Telegram only when both values are set.
 TELEGRAM_BOT_TOKEN=

collector/README.md

@@ -40,6 +40,10 @@ past them.
 Sentinel indexing is committee-scoped and epoch-gated. Unknown coverage gaps
 start a new generation and cannot extend an inactivity streak.
 
+Slash-log chunks are limited to 1,000 blocks. The default provider timeout is
+30 seconds because chunks containing executions require exact archive state
+reconstruction; `L1_SLASH_LOG_PROVIDER_TIMEOUT_MS` may raise it to 45 seconds.
+
 ## Case API
 
 The only API is `/api/v3`:

collector/deploy/slashmon-backend-testing.env.example

@@ -32,6 +32,8 @@ L1_RPC_URL=http://127.0.0.1:8545
 # L1_SLASH_LOG_START_BLOCK=25533241
 # Used only when no exact start block is configured.
 L1_SLASH_LOG_LOOKBACK_BLOCKS=50000
+# Archive calls for a chunk containing slash executions can take longer.
+L1_SLASH_LOG_PROVIDER_TIMEOUT_MS=30000
 
 # Use testing-only notification credentials. In particular, Telegram long
 # polling cannot be shared with production. Leave these empty if not needed.

collector/deploy/slashmon-backend.env.example

@@ -29,6 +29,8 @@ L1_RPC_URL=http://127.0.0.1:8545
 # L1_SLASH_LOG_START_BLOCK=25533241
 # Used only when no exact start block is configured.
 L1_SLASH_LOG_LOOKBACK_BLOCKS=50000
+# Archive calls for a chunk containing slash executions can take longer.
+L1_SLASH_LOG_PROVIDER_TIMEOUT_MS=30000
 
 # Optional notification channels.
 TELEGRAM_BOT_TOKEN=

collector/src/config.mjs

@@ -4,6 +4,8 @@ import { createECDH } from 'node:crypto';
 const DEFAULT_ADMIN_POLL_INTERVAL_MS = 15_000;
 const DEFAULT_SENTINEL_POLL_INTERVAL_MS = 60_000;
 const DEFAULT_L1_POLL_INTERVAL_MS = 30_000;
+const DEFAULT_L1_SLASH_LOG_PROVIDER_TIMEOUT_MS = 30_000;
+const L1_SLASH_LOG_MAX_RUN_MS = 60_000;
 const DEFAULT_DATABASE_PATH = './data/slashmon.sqlite';
 const NETWORK_DEFAULTS = {
   mainnet: {
@@ -117,8 +119,14 @@ export function loadConfig(env = process.env, cwd = process.cwd()) {
     l1SlashLogOverlapBlocks: 12,
     l1SlashLogReorgRewindBlocks: 512,
     l1SlashLogMaxChunksPerPoll: 25,
-    l1SlashLogMaxRunMs: 20_000,
-    l1SlashLogProviderTimeoutMs: 5_000,
+    l1SlashLogMaxRunMs: L1_SLASH_LOG_MAX_RUN_MS,
+    l1SlashLogProviderTimeoutMs: readInteger(
+      env,
+      'L1_SLASH_LOG_PROVIDER_TIMEOUT_MS',
+      DEFAULT_L1_SLASH_LOG_PROVIDER_TIMEOUT_MS,
+      5_000,
+      45_000,
+    ),
 
     bindHost: readString(env.BACKEND_BIND_HOST, '127.0.0.1'),
     port: readInteger(env, 'BACKEND_PORT', 8_790, 1, 65_535),

collector/src/l1-collector.mjs

@@ -8,7 +8,7 @@ export class L1Collector {
     pollIntervalMs,
     maxBackoffMs,
     maxSlashLogChunksPerPoll = 25,
-    maxSlashLogRunMs = 20_000,
+    maxSlashLogRunMs = 60_000,
     logger,
     now = Date.now,
   }) {

collector/src/l1-scanner.mjs

@@ -36,7 +36,7 @@ export class L1Scanner {
     slashLogChunkSize = 1_000,
     slashLogOverlapBlocks = 12,
     slashLogReorgRewindBlocks = 128,
-    slashLogProviderTimeoutMs = 5_000,
+    slashLogProviderTimeoutMs = 30_000,
     clientFactory,
     now = Date.now,
   }) {

collector/test/config.test.mjs

@@ -23,6 +23,8 @@ test('loadConfig provides a complete local configuration', () => {
   assert.equal(config.l1SlashLogStartBlock, undefined);
   assert.equal(config.l1SlashLogLookbackBlocks, 50_000);
   assert.equal(config.l1SlashLogChunkSize, 1_000);
+  assert.equal(config.l1SlashLogProviderTimeoutMs, 30_000);
+  assert.equal(config.l1SlashLogMaxRunMs, 60_000);
   assert.equal(config.sentinelPollIntervalMs, 60_000);
   assert.equal(config.sentinelLookbackEpochs, 3);
   assert.equal(config.sentinelEpochEndBufferSlots, 2);
@@ -90,6 +92,10 @@ test('operator-facing URLs and process settings are validated', () => {
     () => loadConfig({ L1_SLASH_LOG_START_BLOCK: '-1' }),
     /between 0 and 1000000000/,
   );
+  assert.throws(
+    () => loadConfig({ L1_SLASH_LOG_PROVIDER_TIMEOUT_MS: '45001' }),
+    /between 5000 and 45000/,
+  );
   assert.throws(() => loadConfig({ BACKEND_TRUST_PROXY: 'yes' }), /must be true or false/);
   assert.throws(() => loadConfig({ BACKEND_LOG_LEVEL: 'trace' }), /debug, info, warn, or error/);
 
@@ -99,6 +105,7 @@ test('operator-facing URLs and process settings are validated', () => {
     L1_RPC_URL: 'https://rpc-one.example/path',
     L1_SLASH_LOG_START_BLOCK: '25533241',
     L1_SLASH_LOG_LOOKBACK_BLOCKS: '75000',
+    L1_SLASH_LOG_PROVIDER_TIMEOUT_MS: '40000',
     BACKEND_DATABASE_PATH: '../state/slashmon.sqlite',
     BACKEND_PORT: '9000',
     BACKEND_TRUST_PROXY: 'true',
@@ -111,6 +118,7 @@ test('operator-facing URLs and process settings are validated', () => {
   assert.deepEqual(config.l1RpcUrls, ['https://rpc-one.example/path']);
   assert.equal(config.l1SlashLogStartBlock, 25_533_241);
   assert.equal(config.l1SlashLogLookbackBlocks, 75_000);
+  assert.equal(config.l1SlashLogProviderTimeoutMs, 40_000);
   assert.equal(config.databasePath, '/srv/state/slashmon.sqlite');
   assert.equal(config.port, 9000);
   assert.equal(config.trustLoopbackProxy, true);

docs/runbook.md

@@ -32,6 +32,10 @@ database whose durable log cursor was created with a different start, and
 historical observations do not send notifications. Leave it unset to use
 `L1_SLASH_LOG_LOOKBACK_BLOCKS` for a head-relative initial backfill instead.
 Log requests remain fixed at 1,000 blocks to support RPCs with that maximum.
+Each provider gets 30 seconds for a chunk, and each backfill pass yields after
+60 seconds so a fresh L1 snapshot runs regularly. For a slower archive node,
+raise `L1_SLASH_LOG_PROVIDER_TIMEOUT_MS` up to `45000`; keep it below the
+60-second pass budget.
 
 To expand an existing production or testing database, add the setting to that
 service's `/etc/slashmon-backend*.env` and deploy with `--upgrade`. Do not reset