Open_files limit 1024 and network connections timeout (maximum sockets reached?)

Support Kernel
1

Hello, on 5.8.16-2-MANJARO “ulimit -a” shows i had 1024 open files limit, so from root terminal ($ su) i tried “ulimit -n 30000”, it changed the ulimit -a output. The “dstat --socket” command shows that the total sockets is attacking the 1000 mark. The journalctl -r|head -n 40 shows that my proxy software shadowsocks-qt5 continue to flood by messages “ss-qt5[1763811]: TCP connection timeout.”
The server to which this app is connecting has open files limit of 51200. And this timeouts happen also for different server i have tried.

Please can you kindly suggest commands to discover what is the cause or which software limits are reached or how to fix this? Thank you

2

create a file /etc/sysctl.d/20-max-user-watches.conf

fs.inotify.max_user_watches=524288

then reload

sudo sysctl --system

I know nothing of socks

3

Thanks, which effect do you expect? Because i do not know which command or action to do to discover if it helped to something. journalctl ss-qt5 log lines stays same and dstat continue to show what it displayed before.

4

reboot was done
open files (-n) 1024
but “dstat --socket” shows 2k sockets total, where 700 is tcp and 13 udp. :confused:
journalctl continue being flooded by “ss-qt5[1763811]: TCP connection timeout.” kind of notices.

5

17 days later, some people use “ulimit -n 10240” to increase open files limit, but i think it is just temporary and may not be related to network?

6

That’s unrelated to the limit on the number of file descriptors.

The maximum number of open file descriptors has a very significant effect on networking, since all each socket takes up a single file descriptors. In case of TCP, that puts a limit on how many concurrent connections you may have at any given moment.

That will only apply temporarily, in the current process and its children.

7

anyway my other app was complaining about reaching open files limit which is 1024 per “ulimit -n”
Not knowing how this is different from ulimit -n, but system-wide limit of file descriptors is:
$ sysctl fs.file-max
fs.file-max = 9223372036854775807
To change this, it is said to append “fs.file-max = 100000” to sysctl.conf and then “$ sysctl -p”, but “$ locate sysctl.conf” does not look like file is in right path, it is under /etc/ufw/ . The value is in /proc/sys/fs/file-max file (i guess changing the file is temporary effect). Maybe it would work to add new file with 644 permissions and root:root rights and mentioned line to /etc/sysctl.d/ directory.
user wide limit is said to be in:
$ grep -v “#” /etc/security/limits.conf
* hard nofile 31000
i read that the fs.file-max should be number of RAM megabytes:
$ free|grep mem|awk {‘print $2’}
x 64