Replace memcached with redis about zulip HOT 5 OPEN

I quickly tested this and switching the django cache backend "just works". Most of the work here will be packaging django_redis for debian and updating the puppet config. The dev vm will also need it added to the requirements.txt.

pip install django-redis

diff --git a/zproject/settings.py b/zproject/settings.py
index af6341d..aeeeac2 100644
--- a/zproject/settings.py
+++ b/zproject/settings.py
@@ -327,8 +327,8 @@ SESSION_ENGINE = "django.contrib.sessions.backends.cached_db"

 CACHES = {
     'default': {
-        'BACKEND':  'django.core.cache.backends.memcached.PyLibMCCache',
-        'LOCATION': '127.0.0.1:11211',
+        'BACKEND':  'django_redis.cache.RedisCache',
+        'LOCATION': 'redis://127.0.0.1:6379/1',
         'TIMEOUT':  3600
     },
     'database': {

from zulip.

Zulip is using now Redis for the rate limit feature and memcached for the django backend. Do you think it worth consolidating both in a Redis cache, and reduce the number of processes? I can work in the pull request if the team thinks they can approve it.

from zulip.

I'd definitely be interested in a pull request that makes it possible to using redis for the cache; before we switch over I'd like to at least do some measurement of whether there's performance impact since we make extensive use of large multi-get queries (e.g. to fetch 1000 message objects from the cache; see the generic_bulk_cached_fetch in zerver/views/messages.py). A few other details that will need thought:

• scripts/setup/flush-memcached
• redis configuration to manage memory usage etc.

from zulip.

(In particular, I can imagine the Django Python library for redis having a perf issue impacting our usage and I think given an implementation I can fairly easy verify whether that's the case)

from zulip.

It seems both redis and memcached are in use currently (according to the docker-compose file in the docker-zulip project)

from zulip.