I have installed ejabberd 17.06 over Red Hat Enterprise Linux Server release 7.3 (Maipo) with 20GB of HDD and 4GB of RAM
This ejabberd is only used for around 20 users now.
I have a problem. The process "/opt/ejabberd-17.06/bin/beam.smp -K true -P 250000........" has around 80 open files when it starts. If one user log in Jabber, open files increased at 1, and if one user log out, open files decreased at 1. This is normal.
The problem is that anything at server is increasing the open files for the beam.smp process, apart of the login/logout from users.
The ulimit values is 1024 for open files
Many days after the process starts, open files is going to increase until create an error when exceed 1024 open files.
Then, we have this at log:
2017-11-06 21:26:03.322 [error] <0.338.0>@ejabberd_listener:accept:309 (#Port<0.17979>) Failed TCP accept: too many open files
If you check the open files of beam.smp while the error time, is over 1024
If you restart ejabberd, the open files value decreased totally to around 80 again, and after a few days, the problem appears again.
I have config to debug the ejabberd logs, but I don't see useful information of this problem.
It's strange. Is there any
It's strange. Is there any clue in the result of lsof? in my test server with just one client connected: