invoke-ai/InvokeAI · error · ExternalProviderRequestError

DashScope task {task_id} timed out after {_TASK_POLL_TIMEOUT

Error message

DashScope task {task_id} timed out after {_TASK_POLL_TIMEOUT}s

What it means

_poll_task polls the DashScope task status endpoint until SUCCEEDED/FAILED/UNKNOWN, but only up to _TASK_POLL_TIMEOUT seconds. If the task is still pending beyond that, the provider abandons it and raises ExternalProviderRequestError with the task id and timeout length. The remote task may still complete server-side; the client simply stops waiting.

Source

Thrown at invokeai/app/services/external_generation/providers/alibabacloud.py:187

    def _poll_task(
        self,
        base_url: str,
        headers: dict[str, str],
        task_id: str,
        request: ExternalGenerationRequest,
        request_id: str | None,
    ) -> ExternalGenerationResult:
        """Poll an async task until completion."""
        task_url = f"{base_url}/api/v1/tasks/{task_id}"
        start_time = time.monotonic()
        poll_headers = {"Authorization": headers["Authorization"]}
        first_poll = True

        while True:
            elapsed = time.monotonic() - start_time
            if elapsed > _TASK_POLL_TIMEOUT:
                raise ExternalProviderRequestError(f"DashScope task {task_id} timed out after {_TASK_POLL_TIMEOUT}s")

            response = self._get_with_retry(task_url, headers=poll_headers, timeout=30, label="DashScope task poll")
            if not response.ok:
                raise ExternalProviderRequestError(
                    f"DashScope task poll failed with status {response.status_code}: {response.text}"
                )

            data = response.json()
            output = data.get("output", {})
            status = output.get("task_status")

            if first_poll:
                self._logger.info("DashScope task %s submitted (status=%s)", task_id, status)
                first_poll = False

            if status == "SUCCEEDED":
                return self._parse_async_response(output, request, request_id)
            if status in ("FAILED", "UNKNOWN"):

View on GitHub (pinned to 0b6a024f2f)

Solutions

Retry generation, ideally with a smaller/simpler request that finishes faster
Look up the task_id (from logs) directly via the DashScope task query API to see if it later succeeded
Increase _TASK_POLL_TIMEOUT in alibabacloud.py if your workloads legitimately need longer
Check DashScope service status/region health if timeouts are frequent

Example fix

# before (alibabacloud.py)
_TASK_POLL_TIMEOUT = 120
# after
_TASK_POLL_TIMEOUT = 300

Defensive patterns

Strategy: retry

Validate before calling

# estimate feasibility: skip very large requests on slow regions or raise the timeout first
if request.width * request.height > LARGE_IMAGE_THRESHOLD:
    increase_task_poll_timeout()

Try / catch

try:
    result = provider.generate(request)
except ExternalProviderRequestError as e:
    if f'timed out after {_TASK_POLL_TIMEOUT}s' in str(e):
        task_id = extract_task_id_from_logs(e)
        result = query_dashscope_task_once(task_id) or resubmit(request)
    else:
        raise

Prevention

Size _TASK_POLL_TIMEOUT to your largest expected generation
Record task_id (logs) so timed-out tasks can be recovered via the task query API
Prefer smaller/faster models for latency-sensitive flows
Monitor DashScope region health during peak hours

When it happens

Trigger: A DashScope async image task remains in a non-terminal state (PENDING/RUNNING) longer than _TASK_POLL_TIMEOUT seconds — very large images, provider congestion, or a stuck task.

Common situations: Peak-time DashScope queue delays; oversized/complex generation requests; monitoring task_id externally after timeout; polls blocked by network slowness consuming the wall-clock budget.

Understand the failure class

Timeouts: ETIMEDOUT, deadlines, and hung requests — what actually expires when a request times out.

Related errors

AI-assisted analysis of invoke-ai/InvokeAI@0b6a024f2f (2026-08-29). Data as JSON: /api/errors/212684a94731dba2. Report an issue: GitHub.