Why isn't uWSGI respecting the "--http-keepalive" flag?

Question

I installed uWSGI in a Docker container running ubuntu:16.04 using the following commands:

apt-get update
apt-get install -y build-essential python-dev python-pip
pip install uwsgi

I then created a single static file:

cd /root
mkdir static
dd if=/dev/zero bs=1 count=1024 of=static/data

...and finally started uWSGI with the following command:

uwsgi \
    --buffer-size 32768 \
    --http-socket 0.0.0.0:80 \
    --processes 4 \
    --http-timeout 60 \
    --http-keepalive \
    --static-map2=/static=./

I was able to access the static file without any problems. However, despite passing the --http-keepalive option, issuing multiple requests with cURL resulted in the following output:

# curl -v 'http://myserver/static/data' -o /dev/null 'http://myserver/static/data' -o /dev/null
*   Trying 192.168.1.101...
...
> GET /static/data HTTP/1.1
> Host: 192.168.1.101:8100
> User-Agent: curl/7.47.0
> Accept: */*
> 
< HTTP/1.1 200 OK
< Content-Length: 1024
< Last-Modified: Sat, 03 Dec 2016 22:06:49 GMT
< 
{ [1024 bytes data]
100  1024  100  1024    0     0   577k      0 --:--:-- --:--:-- --:--:-- 1000k
* Connection #0 to host 192.168.1.101 left intact
* Found bundle for host 192.168.1.101: 0x563fbc855630 [can pipeline]
* Connection 0 seems to be dead!
* Closing connection 0
...

Of particular interest is the line:

* Connection 0 seems to be dead!

This was confirmed with WireShark:

As you can see, there are two completely separate TCP connections. The first one is closed by uWSGI (packet #10 - the [FIN, ACK]).

What am I doing wrong? Why doesn't uWSGI honor the --http-keepalive flag instead of immediately closing the connection?

Answer 1

In my case, I faced with random 502 responses from aws ALB / ELB. I provided configuration by .ini file like:

http-keepalive = true

but after hours of debugging, I saw a similar picture in Wireshark - after each response, the connection was closed by the server, so the keep-alive option was ignored

In uWSGI#2018, the discussion points that it should be an integer (doc here), but unfortunately can't find exact info on whether it represents seconds of socket lifetime, or it could be simple '1'. After this change - the random 502 disappeared and uwsgi started work in expected mode.

Hope it could be also helpful for somebody.

Answer 2

I was finally able to get keepalive working by switching from --http-socket to simply --http. According to uWSGI docs:

If your web server does not support the uwsgi protocol but is able to speak to upstream HTTP proxies, or if you are using a service like Webfaction or Heroku to host your application, you can use http-socket. If you plan to expose your app to the world with uWSGI only, use the http option instead, as the router/proxy/load-balancer will then be your shield.

In my particular case, it was also necessary to load the http plugin.