Bind health-check listeners on binding_ip and v4v6

[s4heid]

The four health-check listeners (health_check_http_url, its proxy protocol variant, and the per-tcp-proxy health_check_http_tcp-<name> pair) previously bound on :<port>, which is IPv4-wildcard only. On IPv6-only deployments (binding_ip: "::") the listeners were unreachable, causing monit host checks against localhost (IPv4) to fail and restart HAProxy.

Bind them on ha_proxy.binding_ip and append v4v6 like the other frontends so dual-stack works when ha_proxy.v4v6 is enabled.

Behavior change: when binding_ip is set to a specific IPv4 address, the health-check listeners now bind only on that address instead of all IPv4 interfaces. Operators probing health via a different NIC must update their probe targets accordingly.

[peanball]

Looks great. Extends the logic for v4v6 to the health check endpoints, where it was so far missing.

Approving for CI.

[peanball]

v4v6 somewhat negates the binding IP I think. It's only used with the default address (i.e. when it's not explicitly given). Binding to 1.2.3.4:8080 v4v6 will not magically bind an IPv6 socket if I understand the HAProxy docs correctly.

Can you test whether those settings might be mutually exclusive on a real setup, or rather if they have the effect you desire?

Nevermind, the v4v6 flag is only set if the binding ip is :: (i.e. IPv6 default address) and the v4v6 config property is set:

haproxy-boshrelease/jobs/haproxy/templates/haproxy.config.erb

Lines 171 to 178 in f072fbf

	# IPv4 and IPv6 binding (v4v6) Option {{{
	v4v6 = ""
	if_p("ha_proxy.v4v6") do
	if p("ha_proxy.binding_ip") == "::"
	v4v6 = "v4v6"
	end
	end
	# }}}

This is just an extension of what we have in other locations that was just never applied to health checks.

[b1tamara]

Approving to run CI

[peanball]

approving for CI, hopefully one final time :)