Commit 28b2b1d9 authored by Eliot Berriot's avatar Eliot Berriot
Browse files

Merge branch 'reduce-memory-usage' into 'develop'

Added documentation page on how to reduce memory usage.

See merge request funkwhale/funkwhale!177
parents f76f8d53 29645aab
#!/bin/bash -eux
python /app/manage.py collectstatic --noinput
/usr/local/bin/daphne -b 0.0.0.0 -p 5000 config.asgi:application
/usr/local/bin/daphne -b 0.0.0.0 -p 5000 config.asgi:application --proxy-headers
Added documentation for optimizing Funkwhale and reduce its memory
footprint.
Changelog
^^^^^^^^^
For non-docker deployments, add ``--proxy-headers`` at the end of the ``daphne``
command in :file:`/etc/systemd/system/funkwhale-server.service`.
This will ensure the application receive the correct IP address from the client
and not the proxy's one.
......@@ -8,7 +8,7 @@ User=funkwhale
# adapt this depending on the path of your funkwhale installation
WorkingDirectory=/srv/funkwhale/api
EnvironmentFile=/srv/funkwhale/config/.env
ExecStart=/srv/funkwhale/virtualenv/bin/daphne -b ${FUNKWHALE_API_IP} -p ${FUNKWHALE_API_PORT} config.asgi:application
ExecStart=/srv/funkwhale/virtualenv/bin/daphne -b ${FUNKWHALE_API_IP} -p ${FUNKWHALE_API_PORT} config.asgi:application --proxy-headers
[Install]
WantedBy=multi-user.target
......@@ -28,10 +28,16 @@ On a dockerized instance with 2 CPUs and a few active users, the memory footprin
funkwhale_postgres_1 22.73 MiB
funkwhale_redis_1 1.496 MiB
Some users have reported running Funkwhale on Raspberry Pis with a memory
consuption of less than 200MiB.
Thus, Funkwhale should run fine on commodity hardware, small hosting boxes and
Raspberry Pi. We lack real-world exemples of such deployments, so don't hesitate
do give us your feedback (either positive or negative).
Check out :doc:`optimization` for advices on how to tune your instance on small
configurations.
Software requirements
---------------------
......
Optimizing your Funkwhale instance
==================================
Depending on your requirements, you may want to reduce as much as possible
Funkwhale's footprint.
Reduce workers concurrency
--------------------------
Asynchronous tasks are handled by a celery worker, which will by default
spawn a worker process per CPU available. This can lead to a higher
memory usage.
You can control this behaviour using the ``--concurrency`` flag.
For instance, setting ``--concurrency=1`` will spawn only one worker.
This flag should be appended after the ``celery -A funkwhale_api.taskapp worker``
command in your :file:`docker-compose.yml` file if your using Docker, or in your
:file:`/etc/systemd/system/funkwhale-worker.service` otherwise.
.. note::
Reducing concurrency comes at a cost: asynchronous tasks will be processed
more slowly. However, on small instances, this should not be an issue.
Switch from prefork to solo pool
--------------------------------
Using a different pool implementation for Celery tasks may also help.
Using the ``solo`` pool type should reduce your memory consumption.
You can control this behaviour using the ``--pool=solo`` flag.
This flag should be appended after the ``celery -A funkwhale_api.taskapp worker``
command in your :file:`docker-compose.yml` file if your using Docker, or in your
:file:`/etc/systemd/system/funkwhale-worker.service` otherwise.
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment