When does nginx return a Bad gateway (502)?

Server NGINX
1

I have an application behind nginx (Rails on Unicorn if it matters)
which
listens on a UNIX socket.

It all works nice, especially when load is low. When there is some load
on
the server though (say 70%), I randomly(?) get a bunch of 502 responses

usually in batches.

Thus the question: when (under what conditions) does nginx send a 502
response?

Thanks in advance,

fastcatch

PS: The app’s worker processes do not seem to be irresponsive, slow,
blocked
for IO or anything else I have been able to identify as a possible
cause.
Thus I want to understand more the nginx side so as to find the root
cause
– which I believe is on the other side (or some config tweak needed
soemwhere).

Posted at Nginx Forum:

2

In my experience, Nginx returns 502 when the upstream server (Unicorn in
your case) doesn’t respond or terminates the connection unexpectedly.

_Nik

3

Thank you.

BTW: having googled it for different expressions I’ve found that it is a
well known issue and that it has nothing to do with nginx or the Rails
app
itself. Unicorn in and of itself can neither be blamed (albeit IMHO
there
should be some official documentation on this issue, esp. since nginx is
the
preferred/“official” web server for unicorn). It looks like a socket
backlog setting issue to be handled at OS level but I’m still unsure if
that’s a resolution or not.

Posted at Nginx Forum:

4

I can confirm that allowing a longer backlog queue (in unicorn config +
net.core.somaxconn and net.core.netdev_max_backlog) did solve the issue
(for
me).

I posted a bit more details at
Bad gateway errors at load on nginx + Unicorn (Rails 3 app) - Stack Overflow.

Posted at Nginx Forum: