Model timeout failover

[viral pecan]

My agent said this is set to 60 seconds. Is this too low? plugins.entries.codex.config.appServer.requestTimeoutMs = 60000

Mainly use GPT 5.5 on high or x-high and was experiencing a lot of failover when GPT is operating normally and there are no rate limits. Just want to get a second opinion and confirm that this should be higher, from maybe 60 seconds to 300 seconds.

[proven nymph]

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

[timber coral]

Second opinion: yes, 60000 can be too low for Codex app-server control-plane calls, especially with high / x-high reasoning where startup/resume/turn metadata and usage-limit/rate-limit refreshes can be slow.

I’d try 300000 as a reasonable diagnostic value:

plugins: {
  entries: {
    codex: {
      config: {
        appServer: {
          requestTimeoutMs: 300000,
        },
      },
    },
  },
}

[timber coral]

Small nuance: this setting is documented as “Timeout for app-server control-plane calls,” not the whole agent run budget. OpenClaw-owned dynamic tool calls are bounded separately, and the main run also has the agent/session timeout path. So if your logs say codex app-server ... timed out, turn/start timed out, or timeout-shaped model failover, raising this is a good test. If the actual run budget is expiring, also check your agent timeout config.

Can you share openclaw --version too? Version matters here because the Codex harness has been moving quickly.

Docs I checked:
https://docs.openclaw.ai/plugins/codex-harness
https://docs.openclaw.ai/concepts/model-failover
https://docs.openclaw.ai/help/faq

[viral pecan]

I'm using the latest version 5.12