danswer/backend/supervisord.conf

[supervisord]
nodaemon=true
user=root
logfile=/var/log/supervisord.log

# Background jobs that must be run async due to long time to completion
# NOTE: due to an issue with Celery + SQLAlchemy 
# (https://github.com/celery/celery/issues/7007#issuecomment-1740139367)
# we must use the threads pool instead of the default prefork pool for now
# in order to avoid intermittent errors like:
# `billiard.exceptions.WorkerLostError: Worker exited prematurely: signal 11 (SIGSEGV)`. 
#
# This means workers will not be able take advantage of multiple CPU cores 
# on a system, but this should be okay for now since all our celery tasks are
# relatively compute-light (e.g. they tend to just make a bunch of requests to 
# Vespa / Postgres)
[program:celery_worker_primary]
command=celery -A onyx.background.celery.versioned_apps.primary worker 
    --loglevel=INFO 
    --hostname=primary@%%n 
    -Q celery
stdout_logfile=/var/log/celery_worker_primary.log
stdout_logfile_maxbytes=16MB
redirect_stderr=true
autorestart=true
startsecs=10
stopasgroup=true

# NOTE: only allowing configuration here and not in the other celery workers, 
# since this is often the bottleneck for "sync" jobs (e.g. document set syncing, 
# user group syncing, deletion, etc.) 
[program:celery_worker_light]
command=celery -A onyx.background.celery.versioned_apps.light worker
    --loglevel=INFO
    --hostname=light@%%n
    -Q vespa_metadata_sync,connector_deletion,doc_permissions_upsert,checkpoint_cleanup
stdout_logfile=/var/log/celery_worker_light.log
stdout_logfile_maxbytes=16MB
redirect_stderr=true
autorestart=true
startsecs=10
stopasgroup=true

[program:celery_worker_heavy]
command=celery -A onyx.background.celery.versioned_apps.heavy worker 
    --loglevel=INFO 
    --hostname=heavy@%%n
    -Q connector_pruning,connector_doc_permissions_sync,connector_external_group_sync
stdout_logfile=/var/log/celery_worker_heavy.log
stdout_logfile_maxbytes=16MB
redirect_stderr=true
autorestart=true
startsecs=10
stopasgroup=true

[program:celery_worker_indexing]
command=celery -A onyx.background.celery.versioned_apps.indexing worker
    --loglevel=INFO
    --hostname=indexing@%%n
    -Q connector_indexing
stdout_logfile=/var/log/celery_worker_indexing.log
stdout_logfile_maxbytes=16MB
redirect_stderr=true
autorestart=true
startsecs=10
stopasgroup=true

[program:celery_worker_monitoring]
command=celery -A onyx.background.celery.versioned_apps.monitoring worker
    --loglevel=INFO
    --hostname=monitoring@%%n
    -Q monitoring
stdout_logfile=/var/log/celery_worker_monitoring.log
stdout_logfile_maxbytes=16MB
redirect_stderr=true
autorestart=true
startsecs=10
stopasgroup=true

# Job scheduler for periodic tasks
[program:celery_beat]
command=celery -A onyx.background.celery.versioned_apps.beat beat
stdout_logfile=/var/log/celery_beat.log
stdout_logfile_maxbytes=16MB
redirect_stderr=true
startsecs=10
stopasgroup=true

# Listens for Slack messages and responds with answers
# for all channels that the OnyxBot has been added to.
# If not setup, this will just fail 5 times and then stop.
# More details on setup here: https://docs.onyx.app/slack_bot_setup
[program:slack_bot]
command=python onyx/onyxbot/slack/listener.py
stdout_logfile=/var/log/slack_bot.log
stdout_logfile_maxbytes=16MB
redirect_stderr=true
autorestart=true
startretries=5
startsecs=60

# Pushes all logs from the above programs to stdout
# No log rotation here, since it's stdout it's handled by the Docker container logging
[program:log-redirect-handler]
command=tail -qF
    /var/log/celery_beat.log
    /var/log/celery_worker_primary.log
    /var/log/celery_worker_light.log
    /var/log/celery_worker_heavy.log
    /var/log/celery_worker_indexing.log
    /var/log/celery_worker_monitoring.log
    /var/log/slack_bot.log
stdout_logfile=/dev/stdout
stdout_logfile_maxbytes = 0  # must be set to 0 when stdout_logfile=/dev/stdout
autorestart=true