Hello,
During some recent digging around in our uwsgi logs, I noticed that we
often get a few of these messages when we rollout code and restart the
workers:
"*** uWSGI listen queue of socket 3 full !!! (101/100) ***"
Looking around, it became clear that we needed to up the listen queue size
in our config, and I did, setting it to 1024 (and also upping the somaxconn
sysctl variable to 5120 on our machines). Now, when the workers start up,
they say they have a queue of 1024:
"your server socket listen backlog is limited to 1024 connections"
Unfortunately, even though it prints that, it doesn't seem to be taking
that value correctly. We still get error messages exactly like the one
above:
"*** uWSGI listen queue of socket 3 full !!! (101/100) ***"
I'm guessing the second part of "101/100" should be 1024, but it's not
showing up. I should also note that we are running this app in a emperor
setup. Is there something else we need to tune here?
Version of uWSGI: 1.2
Config file (I've hidden some variables, but it's mostly unchanged):
[uwsgi]
uid = user
gid = group
socket = x.x.x.x:xxxx
listen = 1024
workers = 32
max-requests = 20000
harakiri = 60
master = true
single-interpreter = true
virtualenv = env
pythonpath = pythonpath
plugin = python
module = wsgi
pidfile2 = /uwsgi.pid
logto2 = /uwsgi.log
logfile-chmod = 644
stats = x.x.x.x:xxxxx
post-buffering = 65536
buffer-size = 32768
disable-logging = true
chdir = /builds
Let me know if anyone needs anymore info. Thanks for your help!
-Andy
_______________________________________________
uWSGI mailing list
[email protected]
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
