#Performance issue: 5-7 minute response times on Telegram

1 messages · Page 1 of 1 (latest)

pale steppe
#

Running OpenClaw v2026.4.29 on Ubuntu 24.04 (dedicated Hetzner server, 64GB RAM). Single agent on openai-codex/gpt-5.5 via Telegram.
Symptoms:
• Simple “hey” takes 5-7 minutes to get a response
• SessionWriteLockTimeoutError on most messages — lock held 17-29 seconds (max 15s)
• pi-trajectory-flush times out at 10s on every single turn
• active-memory plugin timeoutMs set to 5000 but actual elapsed is 23-38 seconds — timeout not enforced
• fetch-timeout reached on many Codex calls
• Gateway uses 2-3.5GB memory with ~20 running tasks
What I’ve tried:
• Fresh session store (moved old 12MB store to backup)
• Cleared all trajectory files and stale temp files (freed ~1GB)
• Set active-memory model to gemini-2.5-flash-lite with timeoutMs 5000
• Re-authed Codex OAuth
• Network is fine (curl to Google/OpenAI APIs responds in <70ms)
Questions:
1. Is pi-trajectory-flush configurable or disableable? It adds 10s to every turn.
2. Why is plugins.entries.active-memory.config.timeoutMs not being enforced? Config shows 5000 but actual requests run 23-38 seconds.
3. Is there a way to reduce the number of background tasks competing with Telegram? openclaw status shows 20 running / 13,515 tracked tasks.
4. Any known performance issues with the Codex provider holding session write locks during slow model calls?​​​​​​​​​​​​​​​​

slow locust
#

i'm having the same issue

#

it wont respond for minutes, even for commands

grave helm
#

I can't even get OpenClaw to run, even though I followed the installation instructions on the official website.

slow locust
#

there's something wrong with either telegram endpoints or openclaw itself

# 
2026-05-01T15:32:40.448+03:00 error {"subsystem":"gateway/channels/telegram"} telegram deleteWebhook failed: Network request for 'deleteWebhook' failed!
2026-05-01T15:32:50.536+03:00 error {"subsystem":"gateway/channels/telegram"} Telegram webhook cleanup failed: Network request for 'deleteWebhook' failed!; retrying in 2.41s.
2026-05-01T15:32:50.544+03:00 error {"subsystem":"gateway/channels/telegram"} telegram deleteMyCommands failed: Network request for 'deleteMyCommands' failed!
#

constantly gettin these in the logs

bronze shuttle
#

same for me, running on local model + discord. High cpu usage when I send a message - takes minutes before request is send to the server (all of this is after a fresh instal in a fresh ubuntu VM)

molten adder
#

Issue in 4.29:#75591 — Manifest reads 100× per request (90–180s before first token)

  • Root cause: plugin extension manifests aren't cached at startup; they're re-read ~105 times per request, causing 93%
    of CPU time spent waiting on file I/O
  • No working workaround — disabling plugins, wiping caches, reinstalling all failed