一段时间后使用nginx + FastCGI的Django + Satchmo提供HTTP错误502错误的网关

Question

I installed new Satchmo store on Linux Debian 6 distributive. Standard Django development server works perfect, but in production mode with nginx + FastCGI after some time (or it's look like after some memory limit will be overfilled) nginx output "502 Bad gateway" error.

In log files I found these lines:

2011/12/02 02:38:57 [error] 29894#0: *91439 recv() failed (104: Connection reset by peer) while reading response header from upstream, client: 2.95.158.164, server: my-secret-host.com, request: "GET / HTTP/1.1", upstream: "fastcgi://unix:/var/run/www/file.sock:", host: "my-secret-host.com"

I was searching over internet a lot and found nginx can't get right response from my django fastcgi server. I tried different settings of django server (maxchildren, maxrequests), but the error still here (conclusion is higher values - longer life without error). With settings maxchildren=3 maxrequests=10 error randomly appears after 5-10 page refreshes and after 15 refreshes it appears always.

I also found when I comment some lines of satchmo source code error is gone. That's is very strange, because it's very important lines for right working store. I think it can be a clue to the reason of problem. Diff is here: http://dpaste.com/hold/664978/

Problem disappear if I would:

• comment out PAYMENT_PROCESSOR=True line in my only payment module.
• comment out config_register(MultipleStringValue(SHIPPING_GROUP, ...) in shipping/config.py

I think these lines lead to real reasons of crashing my production server. How I can solve this problem completely? Any suggestions to my investigation?

UPDATE:

After enabling satchmo logging I found this message:

Mon, 05 Dec 2011 13:26:37 configuration ERROR Problem finding settings SHOP.SHOW_SITE, server closed the connection unexpectedly This probably means the server terminated abnormally before or while processing the request.

Probably it's possible to solve this using recipe from here https://groups.google.com/group/satchmo-users/browse_thread/thread/506b3ad77e7a766e?hl=es&pli=1 . I will try this a little bit later.

Answer 1

This is caused by a problem with the database connection .

Rationale: The first part of the error message "configuration ERROR Problem finding settings SHOP.SHOW_SITE" is from livesettings. It is usually the first module that uses the database. The second part after comma "server closed the connection unexpectedly..." is a message from Postgres database client about connection to db server.

Database connection, number of active connections and db settings can be tested by following commands:

$ python manage.py dbshell

postgres=> select * from product_product;   -- something typical
postgres=> select * from pg_stat_activity;  -- active connections
postgres=> show all;                        -- show current db server settings

Answer 2

You should take a look at your Satchmo log files and see if there are more details about the reason for this crash. The items you commented out shouldn't prevent a crash, so I suspect there is something else going on with your deployment setup.

Another item to consider, are you using a threaded vs. non-threaded Fast-cgi?