fix(slack): pre-set shuttingDown before app.stop() to prevent orphaned ping intervals by hsiaoa · Pull Request #56646 · openclaw/openclaw

hsiaoa · 2026-03-28T22:43:36Z

Problem

When the health monitor restarts a stale Slack socket-mode connection, a race condition in @slack/[email protected] can leave orphaned ping intervals running indefinitely:

Health monitor detects stale socket → calls stopChannel() → aborts → app.stop()
app.stop() → SocketModeClient.disconnect() sets shuttingDown = true
Race: the library's internal ping interval fires SlackWebSocket.disconnect() → terminate() → emits close BEFORE step 2 sets shuttingDown
The close handler sees shuttingDown === false → schedules delayReconnectAttempt(this.start)
Orphaned connection's ping interval keeps firing, logging "A pong wasn't received..." every ~1.6s and leaking file descriptors

Impact: gateway.err.log grows unbounded (~3 warnings/1.6s per account), FD accumulation over time.

Fix

Introduce gracefulStop() that sets shuttingDown = true on the SocketModeClient BEFORE calling app.stop(), closing the race window. Applied to all three stop paths:

stopOnAbort handler (health monitor restart)
Reconnect loop (disconnect recovery)
finally block (provider teardown)

Why not fix upstream?

@slack/[email protected] has no fix for this (only cosmetic async/await changes). The proper upstream fix would be to clear ping intervals in SocketModeClient.disconnect() before calling websocket.disconnect(), but that requires a library release. This workaround is safe because shuttingDown is a simple boolean sentinel — setting it early just makes the close handler skip the reconnect, which is exactly what we want.

Testing

Verified on macOS with 1 Slack account in socket mode
Health monitor stale-socket restarts no longer produce orphaned pong warnings
gateway.err.log stays clean after restart cycles

AI Disclosure

🤖 AI-assisted (Kiro CLI). Root cause analysis from issue #56508, fix verified against @slack/socket-mode source.

greptile-apps · 2026-03-28T22:46:23Z

Greptile Summary

This PR introduces gracefulStop() to close a race condition in @slack/[email protected] where the library's internal ping timeout can fire disconnect() before SocketModeClient.shuttingDown is set during teardown — causing the close handler to schedule a reconnect attempt and leak ping intervals indefinitely.

The fix is minimal and targeted:

gracefulStop() accesses the receiver's client via a loosely-typed cast and sets shuttingDown = true before delegating to app.stop()
The optional-chaining guard (?.client + if (socketClient)) ensures the mutation is silently skipped in HTTP mode (where HTTPReceiver has no client property), so there is no regression for non-socket deployments
All three teardown paths (stopOnAbort, the reconnect loop, and the finally block) now use gracefulStop() consistently

The approach correctly relies on SocketModeClient.start() resetting shuttingDown on subsequent reconnect attempts (consistent with pre-existing behavior where app.stop() → app.start() cycles already worked), and is appropriately scoped as a workaround pending an upstream library fix.

Confidence Score: 5/5

Safe to merge — the fix is a well-scoped workaround for a confirmed race condition, with no regressions in HTTP mode and consistent teardown behavior across all stop paths.

No P0 or P1 issues found. The private-API access via type cast is intentional and guarded by optional chaining; it degrades silently if the library restructures its internals rather than breaking anything. The .catch(() => undefined) in gracefulStop() preserves the pre-existing error-suppression behavior at all call sites. Testing confirmed reconnect cycles still function correctly, validating that SocketModeClient.start() properly handles a pre-set shuttingDown flag.

No files require special attention.

Important Files Changed

Filename	Overview
extensions/slack/src/monitor/provider.ts	Adds gracefulStop() that pre-sets SocketModeClient.shuttingDown = true before calling app.stop(), closing a race window that caused orphaned ping intervals. Applied consistently to all three stop paths.

_{Reviews (1): Last reviewed commit: "fix(slack): pre-set shuttingDown before ..." | Re-trigger Greptile}

CHE10X · 2026-04-03T15:18:04Z

We're hitting this exact issue in production with 11 Slack accounts configured in socket mode.

Symptoms observed:

360+ pong timeouts per hour
33+ socket reconnects per hour (should be ~1 at startup)
gateway.err.log filling with "A pong wasn't received from the server before the timeout of 5000ms!"
1,400+ unique socket IDs cycling through in a single hour

Environment:

OpenClaw 2026.4.2
11 Slack accounts (separate bots for agent context isolation)
macOS, socket mode

Current workaround: Increased gateway.channelHealthCheckMinutes from 60 to 180 to reduce restart frequency, but this just delays the orphan accumulation.

Would appreciate this fix landing soon — it's a significant stability issue for multi-account Slack deployments. Happy to test a beta if one becomes available.

hsiaoa · 2026-04-03T15:23:47Z

There's no dedicated Slack maintainer yet. We'd appreciate it if @frankekn would help move this along.

…d ping intervals Fixes openclaw#56508 When the health monitor restarts a stale Slack socket-mode connection, there's a race condition in @slack/[email protected]: 1. Health monitor calls stopChannel() → abort signal → app.stop() 2. app.stop() → SocketModeClient.disconnect() sets shuttingDown = true 3. But the library's internal ping interval may fire first, calling SlackWebSocket.disconnect() → terminate() → emits 'close' 4. The close handler checks shuttingDown (still false) and schedules an orphaned reconnect via delayReconnectAttempt(this.start) 5. The orphaned connection's ping interval keeps firing indefinitely, logging 'A pong wasn't received...' every 1.6s and leaking FDs Fix: introduce gracefulStop() that sets shuttingDown = true on the SocketModeClient BEFORE calling app.stop(), closing the race window. Applied to all three stop paths: abort handler, reconnect loop, and finally block. Single file change. No new dependencies.

frankekn · 2026-04-04T12:49:43Z

Merged. Thanks @hsiaoa
Head: f1c91d5

@frankekn

…d ping intervals (openclaw#56646) Merged via squash with admin override. Prepared head SHA: f1c91d5 Note: required red lanes are currently inherited from latest origin/main, not introduced by this PR. Co-authored-by: hsiaoa <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…d ping intervals (openclaw#56646) Merged via squash with admin override. Prepared head SHA: f1c91d5 Note: required red lanes are currently inherited from latest origin/main, not introduced by this PR. Co-authored-by: hsiaoa <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…d ping intervals (openclaw#56646) Merged via squash with admin override. Prepared head SHA: f1c91d5 Note: required red lanes are currently inherited from latest origin/main, not introduced by this PR. Co-authored-by: hsiaoa <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…d ping intervals (openclaw#56646) Merged via squash with admin override. Prepared head SHA: f1c91d5 Note: required red lanes are currently inherited from latest origin/main, not introduced by this PR. Co-authored-by: hsiaoa <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

@frankekn

…d ping intervals (openclaw#56646) Merged via squash with admin override. Prepared head SHA: f1c91d5 Note: required red lanes are currently inherited from latest origin/main, not introduced by this PR. Co-authored-by: hsiaoa <[email protected]> Co-authored-by: frankekn <[email protected]> Reviewed-by: @frankekn

openclaw-barnacle Bot added channel: slack Channel integration: slack size: XS labels Mar 28, 2026

frankekn self-assigned this Apr 4, 2026

openclaw-barnacle Bot added size: S and removed size: XS labels Apr 4, 2026

hsiaoa and others added 2 commits April 4, 2026 20:39

fix: harden Slack socket teardown shutdown

f1c91d5

frankekn force-pushed the fix/slack-socket-orphaned-ping branch from b6c3b14 to f1c91d5 Compare April 4, 2026 12:41

frankekn merged commit ae16452 into openclaw:main Apr 4, 2026
24 of 32 checks passed

jnikolaidis mentioned this pull request May 1, 2026

Slack adapter goes silently dead after channel-stop timeout (post-#56646) #75480

Closed

alexey-pelykh mentioned this pull request May 1, 2026

Slack socket-mode Promise.reject(undefined) on reconnect 408 crashes gateway remoteclaw/remoteclaw#2652

Closed

clawsweeper Bot mentioned this pull request May 5, 2026

Slack provider hardcodes autoReconnectEnabled: false, defeats native socket-mode reconnect #77933

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(slack): pre-set shuttingDown before app.stop() to prevent orphaned ping intervals#56646

fix(slack): pre-set shuttingDown before app.stop() to prevent orphaned ping intervals#56646
frankekn merged 2 commits into
openclaw:mainfrom
hsiaoa:fix/slack-socket-orphaned-ping

hsiaoa commented Mar 28, 2026

Uh oh!

greptile-apps Bot commented Mar 28, 2026

Important Files Changed

Uh oh!

CHE10X commented Apr 3, 2026

Uh oh!

hsiaoa commented Apr 3, 2026

Uh oh!

Uh oh!

frankekn commented Apr 4, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Uh oh!

Conversation

hsiaoa commented Mar 28, 2026

Problem

Fix

Why not fix upstream?

Testing

AI Disclosure

Uh oh!

greptile-apps Bot commented Mar 28, 2026

Greptile Summary

Confidence Score: 5/5

Important Files Changed

Uh oh!

CHE10X commented Apr 3, 2026

Uh oh!

hsiaoa commented Apr 3, 2026

Uh oh!

Uh oh!

frankekn commented Apr 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

frankekn commented Apr 4, 2026 •

edited

Loading