Skip to content
AnglinAI Docs

ERR_MARKER_TIMEOUT — Marker Timeout

HTTP Status: 504 Retryable: Yes Automatic retry: None at the Marker level. The poll loop simply expires after MAX_POLL_TIME_MS (300,000 ms / 5 minutes).

What the User Sees

Document conversion timed out after 5 minutes. (ERR_MARKER_TIMEOUT)

What Causes This Error

The Marker API accepted the PDF for processing but did not complete within the 5-minute poll timeout (MAX_POLL_TIME_MS = 300000). After submission, the system polls the Marker status endpoint at regular intervals. If the status does not reach “complete” within 5 minutes, the system gives up and returns this error.

This most commonly happens with very large PDFs (100+ pages), heavily scanned documents that require OCR, or when Marker’s processing queue is backed up with many concurrent jobs.

The Marker API processes documents asynchronously — the initial submission returns immediately with a job ID, and the system polls until completion. Queue depth on Marker’s side is not visible to us, so during peak times, even moderately sized documents may time out simply because they are waiting in line.

Documents with complex layouts, embedded fonts, or high-resolution images also take longer to process, as Marker performs layout analysis and text extraction on each page.

Resolution Steps

For Users

  1. Try again — the processing queue may have cleared and the document may succeed on retry.
  2. Split the document into smaller sections and convert them individually using page ranges.
  3. Use page ranges to convert the most important sections first.
  4. Very large documents (100+ pages) may consistently time out. Consider splitting the PDF using a tool like Adobe Acrobat or an online PDF splitter.

For Administrators

  1. Check Marker API queue depth if Datalab exposes any metrics or status endpoints.
  2. Consider increasing MAX_POLL_TIME_MS for specific document sizes if timeouts are frequent. This constant is defined in the Marker service module.
  3. Monitor via the structured log event: Marker API processing timed out. Track the frequency and document sizes that trigger timeouts.
  4. If Marker is consistently slow, check if there is a service degradation notice from Datalab.
  5. For documents that consistently time out, consider bypassing Marker and routing through the vision-only conversion pipeline.