Using @cache.action with Redis takes 10 seconds to return a simple HTTP 503 or 404

[Lisandro]

Hi there! I recently upgraded my production environment to web2py 2.16.1, and I'm facing an old issue that was apparently solved.

In my applications I use @cache.action to cache public pages, and as I use Redis for cache, I use it like this:

@cache.action(time_expire=300, cache_model=cache.redis, session=False, vars=False, public=True)
def test():
    ....

Well, I've found that, when the function raises an HTTP 503 or 404, it takes 10 seconds to complete and return the result.

Now, this only happens when cache_model=cache.redis. 

If I change it to cache_model=cache.ram, it takes a few milliseconds to complete as expected.

I'm using Redis 3.2.10 64bit on CentOS7.

The problem had been reported long time ago, and it was also fixed:

https://github.com/web2py/web2py/issues/1355

I've checked my web2py source code, and it's indeed using the latest stable version 2.16.1, and just in case, I looked at gluon/contrib/redis_cache.py and the fix is there.

However, it appears that the problem still can be reproduced with this code:

from gluon.contrib.redis_cache import RedisCache
from gluon.contrib.redis_session import RedisSession
from gluon.contrib.redis_utils import RConn

_redis_conn = RConn('localhost', 6379)
cache.redis = RedisCache(redis_conn=_redis_conn, with_lock=True)

@cache.action(time_expire=300, cache_model=cache.redis, session=False, vars=False, public=True)
def test():
    raise HTTP(503)

Notice that it always takes 10 seconds, maybe that should bring in some clue about what's happening.

Any suggestion?

Thanks in advance!

Regards, Lisandro

[Lisandro]

Quick update: apparently the problem doesn't happen using RedisCache(with_lock=False, ...)

Should this be considered a bug or is it the expected behaviour?

I would like to keep with_lock=True but avoid that 10 second delay when a function raises HTTP 503 or 404.

The issue with that delay is that, during that time, the request is using a database connection, and in my environment I have the risk of exhausting db connections.

I'll try to play a bit with the source code of redis_cache.py, but still any suggestion or comment will be much appreciated.

Thanks!

[Anthony]

I believe this has been fixed in master: https://github.com/web2py/web2py/commit/ea5ea6a30759a2c825f23381540dc396cbc475b7.

Anthony

[Lisandro]

Thank you for your fast answer Anthony!

You are right, the problem is fixed in master.

I applied those changes and the problem was solved.

Thank you very much!