Hi, Simon.
On Tue, Jan 08, 2019 at 10:55:19AM +0100, Simon via Exim-users wrote:
> #exim -d -M 1ggnQv-00010z-Iu
...
> Connecting to smtp-in.orange.fr [80.12.242.9]:25 from 176.31.234.153 ...
> unable to bind outgoing SMTP call to 176.31.234.153: Address already in
> usefailed: Address already in use
It seems that your Exim is configured to use 176.31.234.153 as local
ip-address. To check look for "local_interfaces" variable in config,
or (if you are running Debian) for MAIN_LOCAL_INTERFACES macro and/or
"dc_local_interfaces parameter" in /etc/exim4/update-exim4.conf.conf.
Exim tries to bind() socket to this address before connect() syscall,
and fails if all ports are busy. Address space exhausts much quickly
with bind() than without it. The question is whether you really need
binding to ip=176.31.234.153. If not, you have better to remove it.
On Tue, Jan 08, 2019 at 12:23:46PM +0100, Simon via Exim-users wrote:
> tcp 0 0 176.31.234.153:58282 85.233.202.178:80 TIME_WAIT
> -
> tcp 0 0 176.31.234.153:55682 85.233.202.178:80 TIME_WAIT
> -
> tcp 0 0 176.31.234.153:59792 85.233.202.178:80 TIME_WAIT
> -
> tcp 0 0 176.31.234.153:45427 85.233.202.178:80 TIME_WAIT
> -
> tcp 0 0 176.31.234.153:40023 85.233.202.178:80 TIME_WAIT
> -
>
> lot of line between 25621 !!!!!
Does it mean there are ~26000 similar lines? If so, this is a source of
address space pollution: some application (web client) makes many HTTP
requests to 85.233.202.178, occupying ports on local ip=176.31.234.153.
Quick workaround: change kernel parameter "net.ipv4.ip_local_port_range"
(default is "32768 60999" in modern Linux kernels, for 28230 ports).
Then fix your web application or move it to another host.
--
Eugene Berdnikov
