From 9cc6cdbd5f81e78567886d32bfe66582cb57552f Mon Sep 17 00:00:00 2001 From: Matthias Radestock Date: Mon, 24 Jul 2017 12:28:56 +0100 Subject: [PATCH] increase default conntrack buffer size The figure is based on our experience of running scope probes in production systems. Below this we see a lot of "conntrack stderr:WARNING: We have hit ENOBUFS! We are losing events." --- prog/main.go | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/prog/main.go b/prog/main.go index 52832223dc..82e1a6b987 100644 --- a/prog/main.go +++ b/prog/main.go @@ -288,7 +288,7 @@ func setupFlags(flags *flags) { // Proc & endpoint flag.BoolVar(&flags.probe.useConntrack, "probe.conntrack", true, "also use conntrack to track connections") - flag.IntVar(&flags.probe.conntrackBufferSize, "probe.conntrack.buffersize", 208*1024, "conntrack buffer size") + flag.IntVar(&flags.probe.conntrackBufferSize, "probe.conntrack.buffersize", 4096*1024, "conntrack buffer size") flag.BoolVar(&flags.probe.spyProcs, "probe.proc.spy", true, "associate endpoints with processes (needs root)") flag.StringVar(&flags.probe.procRoot, "probe.proc.root", "/proc", "location of the proc filesystem") flag.BoolVar(&flags.probe.procEnabled, "probe.processes", true, "produce process topology & include procspied connections")