Add netstat_poller #68
Conversation
| fields := map[string]int{ | ||
| "tcp_established": counts["ESTABLISHED"], | ||
| "tcp_syn_sent": counts["SYN_SENT"], | ||
| "tcp_syn_recv": counts["SYN_RECV"], | ||
| "tcp_fin_wait1": counts["FIN_WAIT1"], | ||
| "tcp_fin_wait2": counts["FIN_WAIT2"], | ||
| "tcp_time_wait": counts["TIME_WAIT"], | ||
| "tcp_close": counts["CLOSE"], | ||
| "tcp_close_wait": counts["CLOSE_WAIT"], | ||
| "tcp_last_ack": counts["LAST_ACK"], | ||
| "tcp_listen": counts["LISTEN"], | ||
| "tcp_closing": counts["CLOSING"], | ||
| "tcp_none": counts["NONE"], | ||
| "udp_socket": counts["UDP"], |
There was a problem hiding this comment.
What happens if one of these counts[KEY_NAME] is not present on the map? Are we confident these headings are the same across different distros and/or OSes?
There was a problem hiding this comment.
Good question, this library is providing this interface https://github.com/shirou/gopsutil - even if the value is 0 the count field and label is still there. These are all standard tcp connection state machine states so these should be consistent across all OSes and distros.
There was a problem hiding this comment.
In my opinion, we can focus on linux and leave windows aside. If someone really wants to run shh on windows, they are my guest to fix it if the netstat poller would become a problem.
|
Is the snyk warning due to a GPL component that this PR introduced? |
|
Hi Tilman, no this warning is occurring outside of this PR. I have created another PR to test and confirm which you can see here: https://github.com/heroku/shh/runs/6296402764?check_suite_focus=true |
This PR introduces a new poller which aggregates netstat connection information.