#Gateway restart issues, delays

1 messages · Page 1 of 1 (latest)

manic summit
#

I’m seeing slow Gateway startup / readiness after restart. The HTTP server starts quickly, but Gateway does not become “ready” for ~60-70 seconds. Logs suggest model pricing fetches are timing out during startup.

Recent log excerpt:

2026-04-27T23:30:30.453-06:00 [gateway] starting...
2026-04-27T23:30:35.942-06:00 [gateway] starting HTTP server...
2026-04-27T23:30:36.826-06:00 [gateway] http server listening (8 plugins: acpx, bonjour, browser, device-pair, discord, memory-core, phone-control, talk-voice; 6.4s)
2026-04-27T23:31:36.791-06:00 [model-pricing] OpenRouter pricing fetch failed (timeout 60s): TimeoutError: The operation was aborted due to timeout
2026-04-27T23:31:36.792-06:00 [model-pricing] LiteLLM pricing fetch failed (timeout 60s): TimeoutError: The operation was aborted due to timeout
2026-04-27T23:31:38.459-06:00 [gateway] ready

Default model is openai-codex/gpt-5.5.

Question: is there a supported config option to disable external model pricing fetches, or to seed/override pricing locally so startup does not block on OpenRouter/LiteLLM timeout?

openclaw --version
OpenClaw 2026.4.26 (be8c246)

openclaw gateway status --deep
text
Service: systemd (enabled)
File logs: /tmp/openclaw/openclaw-2026-04-28.log
Command: /usr/bin/node /home/claw/.npm-global/lib/node_modules/openclaw/dist/index.js gateway --port 18789
Service file: ~/.config/systemd/user/openclaw-gateway.service
Service env: OPENCLAW_GATEWAY_PORT=18789

Config (cli): ~/.openclaw/openclaw.json
Config (service): ~/.openclaw/openclaw.json
Gateway: bind=loopback (127.0.0.1), port=18789 (service args)
Probe target: ws://127.0.0.1:18789
Dashboard: http://127.0.0.1:18789/
Probe note: Loopback-only gateway; only local clients can connect.

Runtime: running (pid 27782, state active, sub running, last exit 0, reason 0)
Connectivity probe: ok
Capability: connected-no-operator-scope

Listening: 127.0.0.1:18789
Troubles: run openclaw status
Troubleshooting: https://docs.openclaw.ai/troubleshooting

undone palmBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

brisk troutBOT
#

I checked the 2026.4.26 tag (be8c246): no, not in that release. There isn’t a supported global config toggle there, so adding models.pricing.enabled on 2026.4.26 may not be accepted.

Good news: this was added on main right after that release as:

{
  models: {
    pricing: { enabled: false },
  },
}

That skips the startup OpenRouter/LiteLLM pricing catalog fetches; explicit models.providers.*.models[].cost entries still work for local cost estimates. Docs: https://docs.openclaw.ai/gateway/configuration-reference and https://docs.openclaw.ai/reference/token-use.

I also checked npm: latest is still 2026.4.26, so the clean fix likely needs the next release. Once available:

# 
openclaw config set models.pricing.enabled false
openclaw gateway restart

For 2026.4.26 specifically, the only supported-ish workaround is seeding local cost for every configured model ref so the pricing cache has nothing remote to resolve, but there’s no clean global “don’t fetch pricing” switch until the next release. 🦐

manic summit
#

Is there a proper way to get a message to one of the developers. I had openclaw create a markdown file of what is happening. It did some tests to figure out where the problems are happening and is fairly specific. But it's certainly the type of thing that would need to go to a developer probably.

brisk troutBOT
#

Yep — the proper path is a GitHub issue, not a DM/ping to a developer.

File it here: https://github.com/openclaw/openclaw/issues

For this one, I’d use a title like:

Gateway readiness blocks ~60s on model-pricing catalog timeout during startup

And include:

  • openclaw --version
  • openclaw gateway status --deep
  • the startup log excerpt
  • the markdown report OpenClaw generated
  • what network environment it’s running in, e.g. VPS/home/server/firewall/proxy/DNS restrictions
  • whether OpenRouter/LiteLLM URLs are unreachable from that machine
#

Worth noting in the issue: for 2026.4.26, the released tag doesn’t have the global pricing opt-out yet, but main has since added models.pricing.enabled: false. If your tests found more than that, the markdown report is exactly the kind of thing maintainers need.

If you paste the markdown here, I can help trim it into a clean GitHub issue format. 🦐

manic summit
#

The markdown file is probably too big for a discord post

brisk troutBOT
#

Totally fine — don’t try to cram it into Discord.

Best path:

  1. Open a GitHub issue: https://github.com/openclaw/openclaw/issues
  2. Put a short summary in the issue body.
  3. Attach the markdown report as a file. If GitHub won’t accept .md, rename it to .txt or zip it.
  4. Before attaching, skim for secrets: API keys, tokens, webhook URLs, auth headers, private hostnames if you care.

Suggested issue body:

## Summary
Gateway HTTP server starts quickly, but readiness is delayed ~60–70s because model-pricing catalog fetches time out.

## Version