I've looked at this previous question HAProxy health check and see that the HAProxy directives have changed significantly in this area. The "monitor" directive seems to be the modern way to do this.
I want to have a proxy running in tcp mode, that's capable of reporting its availability to clients.
I can have a separate listener in http mode, that gives a 200OK response:
frontend main
# See "bind" documentation at https://docs.haproxy.org/2.6/configuration.html#4.2-bind
# The proxy will listen on all interfaces for connections to the specified port.
# Connections MUST use the Proxy Protocol (v1 or v2).
# The proxy can ialso Listen on ipv4 and ipv6.
bind :::5000 accept-proxy
bind *:5000 accept-proxy
mode tcp
# Detailed connection logging
log global
option tcplog
# Only certain hosts (sending MTAs) can use this proxy, enforced via ACL
acl valid_client_mta_hosts src 127.0.0.1 172.31.25.101
tcp-request connection reject if !valid_client_mta_hosts
use_backend out
frontend health_check
mode http
bind :::5001
bind *:5001
monitor-uri /haproxy_test
log global # comment this out to omit healthchecks from the logs
however that seems to admit the possibility that 5001 might be up, but there's a problem with 5000.
Is there a way to enable monitoring directly of the mode tcp frontend with recent directives?
Here's a possible workaround:
Use a client that can add the proxy header, to ping the tcp front-end.
Make a request toward the proxy health service. The source and dest of the request can be the "loopback" address.
./happie 35.90.110.253:5000 127.0.0.1:0 127.0.0.1:5001
Sending header version 2
00000000 0d 0a 0d 0a 00 0d 0a 51 55 49 54 0a 21 11 00 0c |.......QUIT.!...|
00000010 7f 00 00 01 7f 00 00 01 00 00 13 89 |............|
HTTP/1.1 200 OK
content-length: 58
cache-control: no-cache
content-type: text/html
<html><body><h1>200 OK</h1>
Service ready.
</body></html>