[Support] Document consumption fails even with OCR_MODE: skip and high memory limits (SIGKILL / WorkerLostError)

[Salvodif]

What's your question or issue?

I am trying to force Paperless-ngx to consume long philosophical texts/books by completely skipping the OCR phase, as the process consistently fails.

Despite configuring the environment to bypass OCR and significantly increasing the available resources, the worker is still being killed with signal 9 (SIGKILL). Even with a 24GB memory limit, the document is never successfully ingested.

System Information
Paperless-ngx version: 2.20.7
Installation method: docker-compose
Host OS: CachyOS
System RAM: 32GB

Current Configuration (docker-compose.env / docker-compose.yml)
I have already applied the following settings to mitigate the issue, but without success:

In docker-compose.env resources

PAPERLESS_OCR_MODE=skip
PAPERLESS_OCR_SKIP_ARCHIVE_FILE=always
PAPERLESS_OCR_CLEAN=none
PAPERLESS_OCR_THREADS=2
PAPERLESS_OCR_PAGES=1

In docker-compose.yml resources

deploy:
  resources:
    limits:
      memory: 24G  # Failed even with this high limit

Additional Context
The issue persists specifically with long documents (books). If I set the memory limit to 4G or even 24G, the consumption process starts but never completes, eventually triggering the SIGKILL.

Since I am already using PAPERLESS_OCR_MODE: skip, why is the worker still crashing? Is there a way to "force-import" a document as a raw binary without any processing/validation that might be causing this OOM-like behavior?

Other logs

[2026-02-27 06:29:57,954] [DEBUG] [celery.beat] beat: Synchronizing schedule...

[2026-02-27 06:29:57,968] [DEBUG] [celery.beat] beat: Waking up in 2.02 seconds.

[2026-02-27 06:30:00,000] [INFO] [celery.beat] Scheduler: Sending due task Check all e-mail accounts (paperless_mail.tasks.process_mail_accounts)

[2026-02-27 06:30:00,007] [DEBUG] [celery.beat] paperless_mail.tasks.process_mail_accounts sent. id->654d61ff-c7a0-44ff-96ac-c9f6cca309bb

[2026-02-27 06:30:00,007] [DEBUG] [celery.beat] beat: Waking up in 5.00 minutes.

[2026-02-27 06:30:00,008] [INFO] [celery.worker.strategy] Task paperless_mail.tasks.process_mail_accounts[654d61ff-c7a0-44ff-96ac-c9f6cca309bb] received

[2026-02-27 06:30:00,009] [DEBUG] [celery.pool] TaskPool: Apply <function fast_trace_task at 0x7ff8b363bec0> (args:('paperless_mail.tasks.process_mail_accounts', '654d61ff-c7a0-44ff-96ac-c9f6cca309bb', {'lang': 'py', 'task': 'paperless_mail.tasks.process_mail_accounts', 'id': '654d61ff-c7a0-44ff-96ac-c9f6cca309bb', 'shadow': None, 'eta': None, 'expires': '2026-02-27T06:39:00.001709+01:00', 'group': None, 'group_index': None, 'retries': 0, 'timelimit': [None, None], 'root_id': '654d61ff-c7a0-44ff-96ac-c9f6cca309bb', 'parent_id': None, 'argsrepr': '()', 'kwargsrepr': '{}', 'origin': 'gen162@3d8966ab4a43', 'ignore_result': False, 'replaced_task_nesting': 0, 'stamped_headers': None, 'stamps': {}, 'properties': {'correlation_id': '654d61ff-c7a0-44ff-96ac-c9f6cca309bb', 'reply_to': '1127ae10-7af7-30f1-91cb-2792cdb804ca', 'delivery_mode': 2, 'expiration': '540000', 'delivery_info': {'exchange': '', 'routing_key': 'celery'}, 'priority': 0, 'body_encoding': 'base64', 'delivery_tag': '92f9ef16-901e-43b1-8e86-6f8880c646b7'}, 'reply_to': '1127ae10-7af7-30f1-91cb-2792cdb804ca', 'correlation_id':... kwargs:{})

[2026-02-27 06:30:47,999] [INFO] [celery.app.trace] Task documents.tasks.index_optimize[ea833d31-e64e-450a-b8a1-0efa00ec22c3] succeeded in 283.88989134199994s: None

[2026-02-27 06:30:48,414] [INFO] [celery.app.trace] Task documents.tasks.empty_trash[1ac9bfaf-0870-42b8-a462-be1dab9b2714] succeeded in 0.07241741699999693s: None

[2026-02-27 06:30:48,746] [INFO] [celery.app.trace] Task paperless_mail.tasks.process_mail_accounts[654d61ff-c7a0-44ff-96ac-c9f6cca309bb] succeeded in 0.029877687999999125s: 'No new documents were added.'

[2026-02-27 06:32:22,186] [INFO] [celery.worker.strategy] Task documents.tasks.consume_file[9bf0838d-eaf9-4f34-9071-dc152fead4e6] received

[2026-02-27 06:32:22,187] [DEBUG] [celery.pool] TaskPool: Apply <function fast_trace_task at 0x7ff8b363bec0> (args:('documents.tasks.consume_file', '9bf0838d-eaf9-4f34-9071-dc152fead4e6', {'lang': 'py', 'task': 'documents.tasks.consume_file', 'id': '9bf0838d-eaf9-4f34-9071-dc152fead4e6', 'shadow': None, 'eta': None, 'expires': None, 'group': None, 'group_index': None, 'retries': 0, 'timelimit': [None, None], 'root_id': '9bf0838d-eaf9-4f34-9071-dc152fead4e6', 'parent_id': None, 'argsrepr': '(ConsumableDocument(source=<DocumentSource.ConsumeFolder: 1>, original_file=PosixPath("/usr/src/paperless/consume/Commentary on Aristotle\'s Nicomachean Ethics (Thomas -- St_ Thomas Aquinas; translated by C_I_ Litzinger; foreword -- Aristotelian.pdf"), original_path=None, mailrule_id=None, mime_type=\'application/pdf\'), DocumentMetadataOverrides(filename=None, title=None, correspondent_id=None, document_type_id=None, tag_ids=None, storage_path_id=None, created=None, asn=None, owner_id=None, view_users=None, view_groups=None, change_users=None,... kwargs:{})

[2026-02-27 06:32:41,070] [ERROR] [celery.worker.request] Task handler raised error: WorkerLostError('Worker exited prematurely: signal 9 (SIGKILL) Job: 9.')

Traceback (most recent call last):

  File "/usr/local/lib/python3.12/site-packages/billiard/pool.py", line 1265, in mark_as_worker_lost

    raise WorkerLostError(

billiard.einfo.ExceptionWithTraceback:

"""

Traceback (most recent call last):

  File "/usr/local/lib/python3.12/site-packages/billiard/pool.py", line 1265, in mark_as_worker_lost

    raise WorkerLostError(

billiard.exceptions.WorkerLostError: Worker exited prematurely: signal 9 (SIGKILL) Job: 9.

"""

What have you tried?

Set PAPERLESS_OCR_MODE: skip.
Limited PAPERLESS_OCR_THREADS to 2.
Increased Docker memory limits from 4GB up to 24GB.
Checked that other smaller PDFs are consumed correctly.

Before give the pdf to paperless I tried to clean it up:

gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dPDFSETTINGS=/prepress -dNOPAUSE -dQUIET -dBATCH -sOutputFile="cassell-gd.pdf" cassell.pdf

and also

TMPDIR=./temp_ocr ocrmypdf --force-ocr --clean --deskew --rotate-pages cassell-gd.pdf cassell-gd-ocr-clean.pdf

Paperless-ngx version

2.20.7

Host OS

CachyOS

Installation method

Docker - official image

System status

No response

Relevant logs or output

Even with OCR_MODE: skip, the logs show a worker crash:

billiard.exceptions.WorkerLostError: Worker exited prematurely: signal 9 (SIGKILL) Job: 61.

[Salvodif]

It seems that it was a problem related to the amount of dedicated RAM.