While running logstash on a remote server from the command line not as a service, I found that when there is a broken pipe, sometimes logstash process is killed and in other cases, the process keeps running on the server.
Broken pipe signal is sent when SSH session is broken while the process is still running.
Expectation: define the behaviour for SIGPIPE and honour it consistently and gracefully.
"consistently" here means: if the behaviour chosen is killing of process, the process should be always killed. Otherwise, if the behaviour of process being intact is chosen, it should always be the behaviour.
Operating System: CentOS 6
ES is configured to listen to LS over network from a box on same subnet and LS in turn gets traffic over network from another box on the same subnet.
Corresponding issue: https://github.com/elastic/logstash/issues/8723