Hi, I have ~6000 hosts sending syslog-TLS to a syslog-ng server. I’m seeing a lot of reconnects from the clients, and on the syslog-ng servers I get the following messages: TCP: request_sock_TCP: Possible SYN flooding on port 6514. Sending cookies. Check SNMP counters. netstat -s show the following: $ netstat -s | grep -i listen 10480620 times the listen queue of a socket overflowed 11144791 SYNs to LISTEN sockets dropped A few resources suggest this could be due to the applications socket listen backlog. Starting syslog-ng with strace reveals the following: 30067 listen(5, 255) = 0 30067 listen(11, 255) = 0 30067 listen(12, 255) = 0 30067 listen(17, 255) = 0 It seems like the backlog is hardcoded to 255 in the source. A bit more info: Saving the output from: “netstat -n | grep -c SYN_RECV” to a file a few times per second, shows the following: 2016-09-19 09:54:14.767469015: 0 2016-09-19 09:54:14.906638364: 0 2016-09-19 09:54:15.048591252: 0 2016-09-19 09:54:15.184325070: 0 2016-09-19 09:54:15.324150368: 0 2016-09-19 09:54:15.459749187: 0 2016-09-19 09:54:15.587983284: 0 2016-09-19 09:54:15.722558975: 42 2016-09-19 09:54:15.873601766: 256 2016-09-19 09:54:16.020145083: 247 2016-09-19 09:54:16.177231109: 178 2016-09-19 09:54:16.340875439: 178 2016-09-19 09:54:16.488506916: 178 2016-09-19 09:54:16.637815500: 176 2016-09-19 09:54:16.790781389: 149 2016-09-19 09:54:16.950254517: 141 2016-09-19 09:54:17.125316109: 131 2016-09-19 09:54:17.286839687: 131 2016-09-19 09:54:17.445583267: 127 2016-09-19 09:54:17.598144758: 127 2016-09-19 09:54:17.747967473: 124 2016-09-19 09:54:17.905359412: 2 2016-09-19 09:54:18.062665358: 2 2016-09-19 09:54:18.239973822: 0 2016-09-19 09:54:18.398021514: 0 2016-09-19 09:54:18.548074304: 0 This matches the entry in /var/log/messages Sep 19 09:54:15 log03 kernel: [1116811.081666] TCP: request_sock_TCP: Possible SYN flooding on port 6514. Sending cookies. Check SNMP counters. Are there any tunings in the syslog-ng configuration I need to implement, or do I have to compile syslog-ng myself with a larger listen backlog? Perhaps this should be configurable? My source definition for syslog-TLS is like this: source s_syslog_tls { syslog( ip(0.0.0.0) port(6514) log_fetch_limit(100) log_iw_size(11000000) max-connections(10000) transport("tls") tls( key-file("/etc/pki/tls/certs/logserver.key") cert-file("/etc/pki/tls/certs/logserver.crt") peer-verify(optional-untrusted) ) ); }; $ syslog-ng -V syslog-ng 3.8.1 Installer-Version: 3.8.1 Revision: Module-Directory: //usr/lib64/syslog-ng Module-Path: //usr/lib64/syslog-ng Available-Modules: disk-buffer,sdjournal,afstomp,json-plugin,cryptofuncs,graphite,cef,kvformat,add-contextual-data,dbparser,pseudofile,curl,csvparser,syslogformat,confgen,afsocket,afuser,date,linux-kmsg-format,system-source,basicfuncs,afamqp,affile,afprog Enable-Debug: off Enable-GProf: off Enable-Memtrace: off Enable-IPv6: on Enable-Spoof-Source: on Enable-TCP-Wrapper: on Enable-Linux-Caps: off syslog-ng-3.8.1-1.el7.centos.x86_64 from https://copr-be.cloud.fedoraproject.org/results/czanik/syslog-ng38/epel-7-x8... Please let me know if more info is needed. /Claus A