t8y2/dbx · error

agent response unavailable: %v

Error message

agent response unavailable: %v

What it means

In agentProcess.call, after writing the JSON-RPC request to the agent's stdin, the harness reads one line from the agent's stdout; if Scan fails there is no response and this error wraps the scanner's Err. It typically means the agent died or hung mid-request.

Source

Thrown at agents/drivers/rabbitmq/bench/agent_compare.go:349

}

func (process *agentProcess) call(method string, params map[string]any) (json.RawMessage, error) {
	process.nextID++
	request := map[string]any{
		"jsonrpc": "2.0",
		"id":      process.nextID,
		"method":  method,
		"params":  params,
	}
	payload, err := json.Marshal(request)
	if err != nil {
		return nil, err
	}
	if _, err := process.stdin.Write(append(payload, '\n')); err != nil {
		return nil, err
	}
	if !process.reader.Scan() {
		return nil, fmt.Errorf("agent response unavailable: %v", process.reader.Err())
	}
	var response agentResponse
	if err := json.Unmarshal(process.reader.Bytes(), &response); err != nil {
		return nil, err
	}
	if response.ID != process.nextID {
		return nil, fmt.Errorf("response id %d does not match request id %d", response.ID, process.nextID)
	}
	if response.Error != nil {
		return nil, errors.New(response.Error.Message)
	}
	return response.Result, nil
}

func (process *agentProcess) close() error {
	_, callError := process.call("shutdown", map[string]any{})
	_ = process.stdin.Close()
	waitError := process.command.Wait()

View on GitHub (pinned to c0390bff16)

Solutions

1. Inspect reader.Err()/agent stderr for the crash reason and fix the underlying agent fault
2. Ensure the broker connection survives the benchmark duration (timeouts, keepalives)
3. Check for OOM/signals (dmesg, ulimit) killing the child under load
4. Add retry/reconnect logic in the agent instead of exiting on transient broker errors

res, err := process.call(method, params)
if err != nil && strings.Contains(err.Error(), "agent response unavailable") {
    // agent died mid-call; capture stderr and exit state
    log.Fatalf("agent unavailable: %v, exit state: %v", err, process.command.ProcessState)
}

When it happens

Trigger: reader.Scan() returns false right after a request was written — the agent exited (crash, OOM kill), its stdout pipe broke, or the agent never answered and the pipe closed.

Common situations: Agent killed by OOM under load; agent crashed on an unhandled method/panic; connection to RabbitMQ dropped and agent exited; benchmark machine resource limits killing the child.

Related errors

• agent exited before {method}
• agent exited before {method}
• timed out during {method}
• {self.candidate.name} {method}: {json.dumps(response['error'
• Custom H2 JDBC JAR does not exist: " + path

AI-assisted analysis of t8y2/dbx@c0390bff16 (2026-09-05). Data as JSON: /api/errors/d3204a64d121f34a. Report an issue: GitHub.