Add TCPTimeouts to netstat default filter

[martinkennelly]

TCP timeouts count is a useful signal to show
abnormal network performance and is another
signal to aid debugging. This metric can be
used to generate proactive alerts for host
network namespace workloads.

Signed-off-by: Martin Kennelly mkennell@redhat.com

[martinkennelly]

I don't understand why test CI is failing.

[ventifus]

You'll want to add this new metric to the end-to-end test's output fixture in collector/fixtures/e2e-output.txt
Here's the change you want to make:

--- collector/fixtures/e2e-output.txt	2021-10-29 15:09:32.563636013 +0000
+++ /tmp/node_exporter_e2e_test.azBPP2/e2e-output.txt	2021-10-29 15:10:43.395416878 +0000
@@ -2240,6 +2240,9 @@
 # HELP node_netstat_TcpExt_SyncookiesSent Statistic TcpExtSyncookiesSent.
 # TYPE node_netstat_TcpExt_SyncookiesSent untyped
 node_netstat_TcpExt_SyncookiesSent 0
+# HELP node_netstat_TcpExt_TCPTimeouts Statistic TcpExtTCPTimeouts.
+# TYPE node_netstat_TcpExt_TCPTimeouts untyped
+node_netstat_TcpExt_TCPTimeouts 115
 # HELP node_netstat_Tcp_ActiveOpens Statistic TcpActiveOpens.
 # TYPE node_netstat_Tcp_ActiveOpens untyped
 node_netstat_Tcp_ActiveOpens 3556

You can check that the end-to-end test passes by running end-to-end-test.sh.

[discordianfish]

Yeah, seems reasonable to me.

[SuperQ]

LGTM, thanks!