t8y2/dbx · error · RuntimeError

RSS probe exited early: {'; '.join(stderr_lines)}

Error message

RSS probe exited early: {'; '.join(stderr_lines)}

What it means

measure_rss() starts the benchmark in 'hold' mode and reads its stdout, waiting for a JSON line with ready=true. If the process's stdout hits EOF while the process has already exited, it raises RuntimeError listing all stderr lines collected, i.e. the probe crashed before signaling readiness.

Source

Thrown at agents/drivers/iotdb/bench/run.py:151

        env=environment | {"IOTDB_BENCH_MODE": "hold"},
        text=True,
        stdout=subprocess.PIPE,
        stderr=subprocess.PIPE,
    )
    stderr_lines: list[str] = []

    def drain_stderr() -> None:
        assert process.stderr is not None
        stderr_lines.extend(line.rstrip() for line in process.stderr)

    threading.Thread(target=drain_stderr, daemon=True).start()
    try:
        assert process.stdout is not None
        deadline = time.monotonic() + env_float("BENCH_READY_TIMEOUT", 30.0)
        while time.monotonic() < deadline:
            line = process.stdout.readline()
            if line == "" and process.poll() is not None:
                raise RuntimeError(f"RSS probe exited early: {'; '.join(stderr_lines)}")
            try:
                payload = json.loads(line)
            except json.JSONDecodeError:
                continue
            if payload.get("ready") is True:
                break
        else:
            raise TimeoutError(f"RSS probe timed out: {'; '.join(stderr_lines)}")
        time.sleep(0.2)
        completed = subprocess.run(
            ["ps", "-o", "rss=", "-p", str(process.pid)],
            text=True,
            stdout=subprocess.PIPE,
            stderr=subprocess.PIPE,
            check=True,
        )
        return float(completed.stdout.strip())
    finally:

View on GitHub (pinned to c0390bff16)

Solutions

1. Read the stderr lines embedded in the RuntimeError for the crash reason
2. Confirm the installed benchmark tool supports IOTDB_BENCH_MODE=hold
3. Run the probe command manually with the same env to see the startup failure
4. Check for port conflicts or missing dependencies that kill the process instantly
5. Verify the probe writes readiness JSON to stdout before doing heavy work

Example fix

// before
$ IOTDB_BENCH_MODE=hold ./bench  # crashes: unknown mode -> RuntimeError: RSS probe exited early
// after
$ upgrade bench tool to a version supporting hold mode, then rerun measure_rss(...)

def preflight_hold_mode(env: dict) -> None:
    probe = subprocess.run(
        ["echo", "check"], env=env | {"IOTDB_BENCH_MODE": "hold"}, capture_output=True, text=True)
    # on the real tool: run with a tiny workload and assert it starts without stderr
    if probe.returncode != 0:
        raise SystemExit(f"probe fails at startup: {probe.stderr}")

try:
    rss = measure_rss(command, environment)
except RuntimeError as e:
    print(f"probe died at startup:\n{e}")  # stderr lines are embedded
    raise SystemExit(1)

When it happens

Trigger: The spawned RSS probe subprocess exits early (non-zero or zero) before emitting {"ready": true}: readline() returns "" and process.poll() is not None. Causes include bad arguments, crash at startup, missing IOTDB_BENCH_MODE support, or instant OOM.

Common situations: Probe binary crashes on startup; env var IOTDB_BENCH_MODE not supported by the installed tool version; port conflict killing the server immediately; stderr shows a Python/JVM traceback.

Related errors

• agent exited before ready
• agent exited during {method}
• agent exited before {method}
• agent exited before {method}
• timed out waiting for agent readiness

AI-assisted analysis of t8y2/dbx@c0390bff16 (2026-09-05). Data as JSON: /api/errors/4a2dd4aad1d11b5e. Report an issue: GitHub.