12

I need to load balance between multiple running servers with different host names. I cannot set-up the same virtual host on each one.

Is it possible to have only one listen configuration with multiple server and make the Health Checks apply the http-send-name-header Host directive? I am using HAProxy 1.5.

I came up with this working haproxy.cfg, as you can see, I had to set a different hostname for each health check as the health check ignores the http-send-name-header Host. I would have preferred to use variables or other methods and keep things more concise.

global
    log 127.0.0.1 local0 notice
    maxconn 2000
    user haproxy
    group haproxy

defaults
    log     global
    mode    http
    option  httplog
    option  dontlognull
    retries 3
    option redispatch
    timeout connect  5000
    timeout client  10000
    timeout server  10000
    stats enable
    stats uri /haproxy?stats
    stats refresh 5s
    balance roundrobin
    option httpclose

listen inbound :80    
    option httpchk HEAD / HTTP/1.1\r\n
    server instance1 127.0.0.101 check inter 3000 fall 1 rise 1
    server instance2 127.0.0.102 check inter 3000 fall 1 rise 1

listen instance1 127.0.0.101:80
    option forwardfor
    http-send-name-header Host
    option httpchk HEAD / HTTP/1.1\r\nHost:\ www.example.com
    server www.example.com www.example.com:80 check inter 5000 fall 3 rise 2

listen instance2 127.0.0.102:80
    option forwardfor
    http-send-name-header Host
    option httpchk HEAD / HTTP/1.1\r\nHost:\ www.bing.com
    server www.bing.com www.bing.com:80 check inter 5000 fall 3 rise 2

2 Answers 2

2
defaults
log global
retries 2
timeout connect 3000
timeout server 5000
timeout client 5000

listen any-name-1
bind IP-Address:port
mode tcp or http
option user-check user haproxy_check
balance roundrobin
server hostname IpAddress:port check
server hostname  IpAddress:port check
listen any-name-2
bind IP-Address:port
mode tcp or http
option user-check user haproxy_check
balance roundrobin
server hostname IpAddress:port check
server hostanme  IpAddress:port check

listen any-name-3
bind IP-Address:port
mode tcp or http
option user-check user haproxy_check
balance roundrobin
server hostname IpAddress:port check
server hostname  IpAddress:port check

listen any-name-4
bind IP-Address:port
mode tcp or http
option user-check user haproxy_check
balance roundrobin
server hostname IpAddress:port check
server hostname  IpAddress:port check

listen any-name-5
bind IP-Address:port
mode tcp or http
option user-check user haproxy_check
balance roundrobin
server hostname IpAddress:port check
server hostname  IpAddress:port check

listen haproxyadmin
bind HAproxyServerIP:HaproxyPort
mode http
stats enable
stats uri /haproxy
stats realm Strictly\ Private
stats auth username:password
1

Update: In the case you described, you need HTTP/1.1 checks which require the host name hardcoded. Given the documentation of version 1.5, there doesn't seem to be a way to avoid this unless you can afford to drop the http checks (which of course is generally not recommended).

Original answer: While I am not familiar with the 1.5 changes of haproxy, what I would do in 1.4 (and I'm fairly sure it still applies in 1.5) is the following. Note that the frontend/backend separation is just a personal convenience and you could just use listen.

defaults
    mode http
    option  httplog
    timeout connect  5000
    timeout client  10000
    timeout server  10000

frontend inbound
    bind 127.0.0.1:8000
    default_backend webservers

backend webservers
    option forwardfor
    option httpchk HEAD / HTTP/1.0
    http-send-name-header Host
    server google www.google.com:80 check inter 5000 fall 3 rise 2
    server bing www.bing.com:80 check inter 5000 fall 3 rise 2

And the result:

$ curl -i localhost:8000
HTTP/1.1 301 Moved Permanently
Cache-Control: no-cache
Content-Length: 0
Location: http://www.bing.com/
Server: Microsoft-IIS/8.0
P3P: CP="NON UNI COM NAV STA LOC CURa DEVa PSAa PSDa OUR IND"
Set-Cookie: _HOP=I=1&TS=1399981378; path=/
Edge-control: no-store
X-MSEdge-Ref: Ref A: 26CEE14531BF45EFAC91FAC3D1945EDF Ref B: 42CE8D142D427C30F7851B56F38837A6 Ref C: Tue May 13 04:42:58 2014 PST
Date: Tue, 13 May 2014 11:42:57 GMT

$ curl -i localhost:8000
HTTP/1.1 301 Moved Permanently
Location: http://www.google.com/
Content-Type: text/html; charset=UTF-8
X-Content-Type-Options: nosniff
Date: Tue, 13 May 2014 11:43:00 GMT
Expires: Thu, 12 Jun 2014 11:43:00 GMT
Cache-Control: public, max-age=2592000
Server: sffe
Content-Length: 219
X-XSS-Protection: 1; mode=block
Alternate-Protocol: 80:quic

<HTML><HEAD><meta http-equiv="content-type" content="text/html;charset=utf-8">
<TITLE>301 Moved</TITLE></HEAD><BODY>
<H1>301 Moved</H1>
The document has moved
<A HREF="http://www.google.com/">here</A>.
</BODY></HTML>
$
4
  • I think that your method would not work as it is not setting the Host header. Without a host header the request will never be reaching the intended web server. You can test this with example.heroku.com May 13, 2014 at 17:37
  • You are right, the httpchk requires the Host header in the heroku example, and there isn't any workaround for it. Dropping the check directive would let the proxy work as intended, but you wouldn't be able to detect in advance a failure.
    – user76776
    May 14, 2014 at 11:12
  • that is exactly my problem, ensuring health check to cloud instances that require the correct host headers. May 14, 2014 at 15:29
  • I cannot drop HTTP/1.1 checks. I will stick with my implementation for now. May 27, 2014 at 9:42

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .