mount() called every request

[citizenfish]

mount() called every request

[adamghill]

I'm curious what you have set up for CACHES in your settings file? Unicorn uses the cache to store component data and I wonder if you have the default local memory storage and that is causing a problem. I use a shared redis as my cache so I've never run into a problem with multiple servers.

[citizenfish]

We don't have any specific setting for CACHES, just using defaults. We were running django using gnuicorn as follows:-

gunicorn --bind 0.0.0.0:80 \
--workers=4 \
--timeout 120 \
--log-level info locaria.wsgi:application \
--access-logfile '-'

We found that reducing the number of workers to 1 fixed the problem but it would be good to understand how we can code round this as running mount every transaction is going to be very inefficient for us.

[adamghill]

Gunicorn workers each have their own memory so the default local memory cache is not shared across workers.

If you only have one server, but multiple workers you could use filesystem caching. If you have multiple servers you will need one of these:

• database caching
• redis caching
• memcache caching

I tend to use redis so in all of my projects I have something like this:

# settings.py

CACHES = {
    "default": {
        "BACKEND": "django.core.cache.backends.redis.RedisCache",
        "LOCATION": "redis://some-redis-server-host:6379",
    }
}

I'm going to look into some potential workarounds to prevent mount from being called, but not having a shared cache will negatively impact other functionality, especially related to child/parent components (see #593 for another example). Unicorn will also be less performant without a shared cache.

Sorry this is not clearly stated in the documentation as a requirement for a production deployment -- I'm going to update the docs to make this more explicit.

[adamghill]

@citizenfish Were you able to use a shared cache to see if this problem went away? Let me know if you still have problems otherwise, I'll close this issue in a few days. Thanks!

[citizenfish]

Yes it did work. Apologies I should have updated the issue. Thanks

[adamghill]

That's great news! I updated https://www.django-unicorn.com/docs/installation/ to mention setting up cache for production deployments. Closing this issue, but let me know if other changes would be helpful.

[Yethiel]

That's great news! I updated https://www.django-unicorn.com/docs/installation/ to mention setting up cache for production deployments. Closing this issue, but let me know if other changes would be helpful.

The docs don't seem to mention it yet.
I ran into the same issue: Locally everything works fine but on production everything was very buggy and unpredictable.

I'm glad it's only a cache issue, unicorn has been great so far! I enabled file based caching and now it works beautifully.