100% CPU after upgrade to consul 1.5.2

[42wim]

After upgrading to 1.5.2 our consul agents where taking all available CPU.

I traced the issue to PR #5468

We specify our IPv6 recursor on startup with the flag -recursor [2001:db8:85a3:8d3:1319:8a2e:370:7348] but without a port (which should use the default port 53)

This PR introduced a tight loop on that port 53 check in:

consul/agent/dns.go

Lines 271 to 275 in a82e6a7

	START:
	_, _, err := net.SplitHostPort(recursor)
	if ae, ok := err.(*net.AddrError); ok && ae.Err == "missing port in address" {
	recursor = ipaddr.FormatAddressPort(recursor, 53)
	goto START

which results in an every increasing bracket adding [[[2001:db8:85a3:8d3:1319:8a2e:370:7348]]:53]:53 etc and taking all the CPU

According to golang/go#20470 the "[]" are only allowed when specifying a port with an ipv6 address.

But the -recursor 2001:db8:85a3:8d3:1319:8a2e:370:7348 fails with too many colons in address

So a possible fix would be to check for that error as below:

START:
	_, _, err := net.SplitHostPort(recursor)
	if ae, ok := err.(*net.AddrError); ok && (ae.Err == "missing port in address" ||  ae.Err == "too many colons in address") {
		recursor = ipaddr.FormatAddressPort(recursor, 53)
		goto START

ping @pierresouchay

Simple workaround for this bug is specifying the port -recursor [2001:db8:85a3:8d3:1319:8a2e:370:7348]:53

[banks]

Thanks @42wim for the great debugging tips too.

You're proposed fix sounds like a good one to me. If you'd like to contribute it as a PR with a test that would be much appreciated. But if not we will take a look.

[42wim]

Made PR for this in #6128

[stale]

Hey there, This issue has been automatically closed because there hasn't been any activity for at least 90 days. If you are still experiencing problems, or still have questions, feel free to open a new one 👍