How to set unix socket backlog with systemd?

Question

I've got a problem with nginx and a fastcgi backend. I've tried to increase the backlog size but it doesn't appear to be working. What am I doing wrong? The backend is started via systemd.

OS: 16.04.2 LTS (Xenial Xerus)

# /etc/systemd/system/backend.socket 
[Socket]
Backlog=1000
ListenStream=/tmp/backend.socket

# ss -l|grep back
u_str  LISTEN     0      128    /tmp/backend.socket 21641                 * 0             

# nginx log
2017/06/21 09:40:56 [error] 1565#1565: *33729 connect() to unix:/tmp/backend.socket failed (11: Resource temporarily unavailable) while connecting to upstream, upstream: "fastcgi://unix:/tmp/backend.socket:"

$ cat /proc/sys/net/core/somaxconn
1024

score 1 · Answer 1 · answered Jun 21 '17 at 20:52

1

From listen(2):

If the backlog argument is greater than the value in /proc/sys/net/core/somaxconn, then it is silently truncated to that value; the default value in this file is 128. In kernels before 2.4.25, this limit was a hard coded value, SOMAXCONN, with the value 128.

So, you need to bump the value in /proc/sys/net/core/somaxconn to your desired backlog length.

answered Jun 21 '17 at 20:52

Tollef Fog Heen

692
3
10

$ cat /proc/sys/net/core/somaxconn 1024 – XTF Jun 22 '17 at 10:04
And you changed this before the socket was set up/started? – Tollef Fog Heen Jun 22 '17 at 14:13
Just rebooted entire VM just to be sure.. – XTF Jun 22 '17 at 14:58

score 0 · Accepted Answer · answered Jun 25 '17 at 15:19

The Backlog parameter was correct. The problem was, the .socket file was not enabled via systemctl enable x.socket, so systemd was only starting the daemon and the daemon itself was creating the socket, obviously not taking the backlog parameter in the .socket file into account.

How to set unix socket backlog with systemd?

2 Answers2

Linked