Question

I can’t seem to get my @postfork functions to run…

  1 import uwsgi
  2 from uwsgidecorators import *
  3 from gevent import monkey; monkey.patch_all()
  4 import sys
  5 import umysql
  6 import time
  7 
  8 DB_HOST = 'stage.masked.com'
  9 DB_PORT = 3306
 10 DB_USER = 'masked'
 11 DB_PASSWD = 'masked'
 12 DB_DB = 'masked'
 13 
 14 mysql_conn = None
 15 
 16 @postfork
 17 def zebra():
 18     print "I AM ZEBRA"
 19     raise
 20     
 21 @postfork
 22 def setup_pool():
 23     global mysql_conn
 24     mysql_conn = umysql.Connection()
 25     print "HIII" 
 26     sys.stderr.write(str(mysql_conn.is_connected()))
 27     mysql_conn.connect (DB_HOST, DB_PORT, DB_USER, DB_PASSWD, DB_DB)
 28     sys.stderr.write(str(mysql_conn.is_connected()))
 29     
 30 def application(env, start_response):
 31     print "HALLO"

When I start uwsgi, I get nothing until I hit the route (defined in nginx that calls this py app). Here’s how I start uwsgi:

# uwsgi --socket=/tmp/uwsgi_server.sock --master --processes=2 --listen=4096 --disable-logging --loop gevent --async 128 --uid=www-data --gid=www-data --vhost
*** Starting uWSGI 1.1.2 (64bit) on [Thu Apr 19 23:55:32 2012] ***
compiled with version: 4.6.1 on 18 April 2012 20:10:46
current working directory: /ebs/py
detected binary path: /usr/local/bin/uwsgi
setgid() to 33
setuid() to 33
your memory page size is 4096 bytes
detected max file descriptor number: 1024
async fd table size: 1024
allocated 130048 bytes (127 KB) for 128 cores per worker.
VirtualHosting mode enabled.
lock engine: pthread mutexes
uwsgi socket 0 bound to UNIX address /tmp/uwsgi_server.sock fd 3
Python version: 2.7.2+ (default, Oct  4 2011, 20:41:12)  [GCC 4.6.1]
Python main interpreter initialized at 0xbb9ad0
your server socket listen backlog is limited to 4096 connections
*** Operational MODE: preforking+async ***
*** no app loaded. going in full dynamic mode ***
*** uWSGI is running in multiple interpreter mode ***
spawned uWSGI master process (pid: 3340)
spawned uWSGI worker 1 (pid: 3341, cores: 128)
spawned uWSGI worker 2 (pid: 3342, cores: 128)
*** running gevent loop engine [addr:0x451080] ***
WSGI app 0 (mountpoint='masked.com|') ready in 0 seconds on interpreter 0xbb9ad0 pid: 3342

When I hit the route, I get:

HALLO

How can I get my @postfork functions to run? My end goal is to get connection pooling in the application function.

Thanks!

UPDATE: If I swap out –vhost for –wsgi-file=server.py, it works as expected.

Answer 1

there is no catchable fork() if you use dynamic apps (the fork() all happen before app loading)

You can move postfork hooks in a .py file you can import on server startup with

–import modulename/filename