Dear list,

I'm getting intermittent 502 errors when I use "cheaper" or "max-requests"
with Nginx and uWSGI. About 1 in 200 queries fail.

Here's the log from Nginx :
2019/08/16 10:17:24 [error] 24#24: *75032 upstream prematurely closed
connection while reading response header from upstream, client:, server:, request: "GET
/mobile/groupe_liste/ HTTP/1.1", upstream: "uwsgi://", host:
"", referrer: "";

It seems these happen when workers are killed, as at the same timestamp, in
the uWSGI  log, I have :
worker 2 killed successfully (pid: 1350)
uWSGI worker 2 cheaped.

When I run without cheaper or max-requests (i.e. when I launch
uWSGI without --cheaper=1 --cheaper-step=4 --max-requests=100), I get no
errors. Obviously I don't want to do this, as my app has some long periods
with no activity, and short periods with heavy activity, and I'm trying to
avoid wasting too much memory on idle workers.

I'm using uWSGI 2.0.18 and Nginx 1.13.7, in two separate Docker containers

Here's the whole command I use to run uWSGI that has issues :
uwsgi --module=myapp.wsgi --need-app --socket=:8000 --workers=20
--threads=4 --buffer-size=8192 --cheaper=1 --cheaper-step=4

The one without issues :
uwsgi --module=myapp.wsgi --need-app --socket=:8000 --workers=20
--threads=4 --buffer-size=8192

And here's the config used in Nginx :
location / {
uwsgi_pass django:8000;
uwsgi_read_timeout 300;
include /etc/nginx/uwsgi_params;

I googled extensively, found somme issues [1] [2] that seemed related, but
comments didn't help.

And idea on why this is happening ? What else can I try to get this working
properly ?

Thanks !!


[1] :
[2] :
uWSGI mailing list

Reply via email to