We are often seeing the following error messages from HEC servers and users are complaining of failures connecting to HEC:
04-16-2020 19:02:04.513 +0000 WARN HttpListener - Socket error from 10.1.32.176:3655 while accessing /services/collector/event/1.0: Connection reset by peer
04-16-2020 19:21:51.387 +0000 WARN HttpListener - HTTP active connections down to 1354, unthrottling
04-16-2020 19:21:51.354 +0000 WARN HttpListener - 1365 HTTP connections active, throttling
The number 1365 is limited by the open files, 4096 in ulimit ( 3 X 1365 = 4095) . To avoid the connectivity issue itself you can increase the open files by modifying the etc/security/limits.conf in Linux. If you find it is there already then consider to add it to splunk start up script
i) initd script, add the below and change the start/restart to call the function
set_ulimits() {
ulimit -Hn 65535
ulimit -Sn 65535
}
start)
set_ulimits
splunk_start
;;
restart)
set_ulimits
splunk_restart
ii) Under systemd you can increase it using LimitNOFILE
[Service]
Type=simple
Restart=always
ExecStart=/home/splunk/s734/bin/splunk _internal_launch_under_systemd
LimitNOFILE=655360
But 1300+ HEC connections are very high you may want to monitor the performance over the network, like netstat -an will show if the TCP Recv queue is building up or not. Then you may need to load balance the # of connections by spawning another HEC receiver.
The number 1365 is limited by the open files, 4096 in ulimit ( 3 X 1365 = 4095) . To avoid the connectivity issue itself you can increase the open files by modifying the etc/security/limits.conf in Linux. If you find it is there already then consider to add it to splunk start up script
i) initd script, add the below and change the start/restart to call the function
set_ulimits() {
ulimit -Hn 65535
ulimit -Sn 65535
}
start)
set_ulimits
splunk_start
;;
restart)
set_ulimits
splunk_restart
ii) Under systemd you can increase it using LimitNOFILE
[Service]
Type=simple
Restart=always
ExecStart=/home/splunk/s734/bin/splunk _internal_launch_under_systemd
LimitNOFILE=655360
But 1300+ HEC connections are very high you may want to monitor the performance over the network, like netstat -an will show if the TCP Recv queue is building up or not. Then you may need to load balance the # of connections by spawning another HEC receiver.