Archive for the ‘Haproxy’ Category

Webserver farm behind Load Balancer Proxy or how to preserve incoming internet IP to local net IP Apache webservers by adding additional haproxy header with remoteip

Monday, April 18th, 2022

logo-haproxy-apache-remoteip-configure-and-check-to-have-logged-real-ip-address-inside-apache-forwarded-from-load-balancer

Having a Proxy server for Load Balancing is a common solutions to assure High Availability of Web Application service behind a proxy.
You can have for example 1 Apache HTTPD webservers serving traffic Actively on one Location (i.e. one city or Country) and 3 configured in the F5 LB or haproxy to silently keep up and wait for incoming connections as an (Active Failure) Backup solution

Lets say the Webservers usually are set to have local class C IPs as 192.168.0.XXX or 10.10.10.XXX and living in isolated DMZed well firewalled LAN network and Haproxy is configured to receive traffic via a Internet IP 109.104.212.13 address and send the traffic in mode tcp via a NATTed connection (e.g. due to the network address translation the source IP of the incoming connections from Intenet clients appears as the NATTed IP 192.168.1.50.

The result is that all incoming connections from haproxy -> webservers will be logged in Webservers /var/log/apache2/access.log wrongly as incoming from source IP: 192.168.1.50, meaning all the information on the source Internet Real IP gets lost.

load-balancer-high-availailibility-haproxy-apache
 

How to pass Real (Internet) Source IPs from Haproxy "mode tcp" to Local LAN Webservers  ?
 

Usually the normal way to work around this with Apache Reverse Proxies configured is to use HTTP_X_FORWARDED_FOR variable in haproxy when using HTTP traffic application that is proxied (.e.g haproxy.cfg has mode http configured), you have to add to listen listener_name directive or frontend Frontend_of_proxy

option forwardfor
option http-server-close

However unfortunately, IP Header preservation with X_FORWADED_FOR  HTTP-Header is not possible when haproxy is configured to forward traffic using mode tcp.

Thus when you're forced to use mode tcp to completely pass any traffic incoming to Haproxy from itself to End side, the solution is to
 

  • Use mod_remoteip infamous module that is part of standard Apache installs both on apache2 installed from (.deb) package  or httpd rpm (on redhats / centos).

 

1. Configure Haproxies to send received connects as send-proxy traffic

 

The idea is very simple all the received requests from outside clients to Haproxy are to be send via the haproxy to the webserver in a PROXY protocol string, this is done via send-proxy

             send-proxy  – send a PROXY protocol string

Rawly my current /etc/haproxy/haproxy.cfg looks like this:
 

global
        log /dev/log    local0
        log /dev/log    local1 notice
        chroot /var/lib/haproxy
        user haproxy
        group haproxy
        daemon
        maxconn 99999
        nbproc          1
        nbthread 2
        cpu-map         1 0
        cpu-map         2 1


defaults
        log     global
       mode    tcp


        timeout connect 5000
        timeout connect 30s
        timeout server 10s

    timeout queue 5s
    timeout tunnel 2m
    timeout client-fin 1s
    timeout server-fin 1s

                option forwardfor

    retries                 15

 

 

frontend http-in
                mode tcp

                option tcplog
        log global

                option logasap
                option forwardfor
                bind 109.104.212.130:80
    fullconn 20000
default_backend http-websrv
backend http-websrv
        balance source
                maxconn 3000

stick match src
    stick-table type ip size 200k expire 30m
        stick on src


        server ha1server-1 192.168.0.205:80 check send-proxy weight 254 backup
        server ha1server-2 192.168.1.15:80 check send-proxy weight 255
        server ha1server-3 192.168.2.30:80 check send-proxy weight 252 backup
        server ha1server-4 192.168.1.198:80 check send-proxy weight 253 backup
                server ha1server-5 192.168.0.1:80 maxconn 3000 check send-proxy weight 251 backup

 

 

frontend https-in
                mode tcp

                option tcplog
                log global

                option logasap
                option forwardfor
        maxconn 99999
           bind 109.104.212.130:443
        default_backend https-websrv
                backend https-websrv
        balance source
                maxconn 3000
        stick on src
    stick-table type ip size 200k expire 30m


                server ha1server-1 192.168.0.205:443 maxconn 8000 check send-proxy weight 254 backup
                server ha1server-2 192.168.1.15:443 maxconn 10000 check send-proxy weight 255
        server ha1server-3 192.168.2.30:443 maxconn 8000 check send-proxy weight 252 backup
        server ha1server-4 192.168.1.198:443 maxconn 10000 check send-proxy weight 253 backup
                server ha1server-5 192.168.0.1:443 maxconn 3000 check send-proxy weight 251 backup

listen stats
    mode http
    option httplog
    option http-server-close
    maxconn 10
    stats enable
    stats show-legends
    stats refresh 5s
    stats realm Haproxy\ Statistics
    stats admin if TRUE

 

After preparing your haproxy.cfg and reloading haproxy in /var/log/haproxy.log you should have the Real Source IPs logged in:
 

root@webserver:~# tail -n 10 /var/log/haproxy.log
Apr 15 22:47:34 pcfr_hware_local_ip haproxy[2914]: 159.223.65.16:58735 [15/Apr/2022:22:47:34.586] https-in https-websrv/ha1server-2 1/0/+0 +0 — 7/7/7/7/0 0/0
Apr 15 22:47:34 pcfr_hware_local_ip haproxy[2914]: 20.113.133.8:56405 [15/Apr/2022:22:47:34.744] https-in https-websrv/ha1server-2 1/0/+0 +0 — 7/7/7/7/0 0/0
Apr 15 22:47:35 pcfr_hware_local_ip haproxy[2914]: 54.36.148.248:15653 [15/Apr/2022:22:47:35.057] https-in https-websrv/ha1server-2 1/0/+0 +0 — 7/7/7/7/0 0/0
Apr 15 22:47:35 pcfr_hware_local_ip haproxy[2914]: 185.191.171.35:26564 [15/Apr/2022:22:47:35.071] https-in https-websrv/ha1server-2 1/0/+0 +0 — 8/8/8/8/0 0/0
Apr 15 22:47:35 pcfr_hware_local_ip haproxy[2914]: 213.183.53.58:42984 [15/Apr/2022:22:47:35.669] https-in https-websrv/ha1server-2 1/0/+0 +0 — 6/6/6/6/0 0/0
Apr 15 22:47:35 pcfr_hware_local_ip haproxy[2914]: 159.223.65.16:54006 [15/Apr/2022:22:47:35.703] https-in https-websrv/ha1server-2 1/0/+0 +0 — 7/7/7/7/0 0/0
Apr 15 22:47:36 pcfr_hware_local_ip haproxy[2914]: 192.241.113.203:30877 [15/Apr/2022:22:47:36.651] https-in https-websrv/ha1server-2 1/0/+0 +0 — 4/4/4/4/0 0/0
Apr 15 22:47:36 pcfr_hware_local_ip haproxy[2914]: 185.191.171.9:6776 [15/Apr/2022:22:47:36.683] https-in https-websrv/ha1server-2 1/0/+0 +0 — 5/5/5/5/0 0/0
Apr 15 22:47:36 pcfr_hware_local_ip haproxy[2914]: 159.223.65.16:64310 [15/Apr/2022:22:47:36.797] https-in https-websrv/ha1server-2 1/0/+0 +0 — 6/6/6/6/0 0/0
Apr 15 22:47:36 pcfr_hware_local_ip haproxy[2914]: 185.191.171.3:23364 [15/Apr/2022:22:47:36.834] https-in https-websrv/ha1server-2 1/1/+1 +0 — 7/7/7/7/0 0/0

 

2. Enable remoteip proxy protocol on Webservers

Login to each Apache HTTPD and to enable remoteip module run:
 

# a2enmod remoteip


On Debians, the command should produce a right symlink to mods-enabled/ directory
 

# ls -al /etc/apache2/mods-enabled/*remote*
lrwxrwxrwx 1 root root 31 Mar 30  2021 /etc/apache2/mods-enabled/remoteip.load -> ../mods-available/remoteip.load

 

3. Modify remoteip.conf file and allow IPs of haproxies or F5s

 

Configure RemoteIPTrustedProxy for every Source IP of haproxy to allow it to send X-Forwarded-For header to Apache,

Here are few examples, from my apache working config on Debian 11.2 (Bullseye):
 

webserver:~# cat remoteip.conf
RemoteIPHeader X-Forwarded-For
RemoteIPTrustedProxy 192.168.0.1
RemoteIPTrustedProxy 192.168.0.205
RemoteIPTrustedProxy 192.168.1.15
RemoteIPTrustedProxy 192.168.0.198
RemoteIPTrustedProxy 192.168.2.33
RemoteIPTrustedProxy 192.168.2.30
RemoteIPTrustedProxy 192.168.0.215
#RemoteIPTrustedProxy 51.89.232.41

On RedHat / Fedora other RPM based Linux distrubutions, you can do the same by including inside httpd.conf or virtualhost configuration something like:
 

<IfModule remoteip_module>
      RemoteIPHeader X-Forwarded-For
      RemoteIPInternalProxy 192.168.0.0/16
      RemoteIPTrustedProxy 192.168.0.215/32
</IfModule>


4. Enable RemoteIP Proxy Protocol in apache2.conf / httpd.conf or Virtualhost custom config
 

Modify both haproxy / haproxies config as well as enable the RemoteIP module on Apache webservers (VirtualHosts if such used) and either in <VirtualHost> block or in main http config include:

RemoteIPProxyProtocol On


5. Change default configured Apache LogFormat

In Domain Vhost or apache2.conf / httpd.conf

Default logging Format will be something like:
 

LogFormat "%h %l %u %t \"%r\" %>s %O \"%{Referer}i\" \"%{User-Agent}i\"" combined


or
 

LogFormat "%v:%p %h %l %u %t \"%r\" %>s %O \"%{Referer}i\" \"%{User-Agent}i\"" combined

 

Once you find it in /etc/apache2/apache2.conf / httpd.conf or Vhost, you have to comment out this by adding shebang infont of sentence make it look as follows:
 

LogFormat "%v:%p %a %l %u %t \"%r\" %>s %O \"%{Referer}i\" \"%{User-Agent}i\"" vhost_combined
LogFormat "%a %l %u %t \"%r\" %>s %O \"%{Referer}i\" \"%{User-Agent}i\"" combined
LogFormat "%a %l %u %t \"%r\" %>s %O" common
LogFormat "%{Referer}i -> %U" referer
LogFormat "%{User-agent}i" agent


The Changed LogFormat instructs Apache to log the client IP as recorded by mod_remoteip (%a) rather than hostname (%h). For a full explanation of all the options check the official HTTP Server documentation page apache_mod_config on Custom Log Formats.

and reload each Apache server.

on Debian:

# apache2ctl -k reload

On CentOS

# systemctl restart httpd


6. Check proxy protocol is properly enabled on Apaches

 

remoteip module will enable Apache to expect a proxy connect header passed to it otherwise it will respond with Bad Request, because it will detect a plain HTML request instead of Proxy Protocol CONNECT, here is the usual telnet test to fetch the index.htm page.

root@webserver:~# telnet localhost 80
Trying 127.0.0.1…
Connected to localhost.
Escape character is '^]'.
GET / HTTP/1.1

HTTP/1.1 400 Bad Request
Date: Fri, 15 Apr 2022 19:04:51 GMT
Server: Apache/2.4.51 (Debian)
Content-Length: 312
Connection: close
Content-Type: text/html; charset=iso-8859-1

<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2.0//EN">
<html><head>
<title>400 Bad Request</title>
</head><body>
<h1>Bad Request</h1>
<p>Your browser sent a request that this server could not understand.<br />
</p>
<hr>
<address>Apache/2.4.51 (Debian) Server at grafana.pc-freak.net Port 80</address>
</body></html>
Connection closed by foreign host.

 

root@webserver:~# telnet localhost 80
Trying 127.0.0.1…
Connected to localhost.
Escape character is '^]'.
HEAD / HTTP/1.1

HTTP/1.1 400 Bad Request
Date: Fri, 15 Apr 2022 19:05:07 GMT
Server: Apache/2.4.51 (Debian)
Connection: close
Content-Type: text/html; charset=iso-8859-1

Connection closed by foreign host.


To test it with telnet you can follow the Proxy CONNECT syntax and simulate you're connecting from a proxy server, like that:
 

root@webserver:~# telnet localhost 80
Trying 127.0.0.1…
Connected to localhost.
Escape character is '^]'.
CONNECT localhost:80 HTTP/1.0

HTTP/1.1 301 Moved Permanently
Date: Fri, 15 Apr 2022 19:13:38 GMT
Server: Apache/2.4.51 (Debian)
Location: https://zabbix.pc-freak.net
Cache-Control: max-age=900
Expires: Fri, 15 Apr 2022 19:28:38 GMT
Content-Length: 310
Connection: close
Content-Type: text/html; charset=iso-8859-1

<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2.0//EN">
<html><head>
<title>301 Moved Permanently</title>
</head><body>
<h1>Moved Permanently</h1>
<p>The document has moved <a href="https://zabbix.pc-freak.net">here</a>.</p>
<hr>
<address>Apache/2.4.51 (Debian) Server at localhost Port 80</address>
</body></html>
Connection closed by foreign host.

You can test with curl simulating the proxy protocol CONNECT with:

root@webserver:~# curl –insecure –haproxy-protocol https://192.168.2.30

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"
"http://w3.org/TR/html4/loose.dtd">
<html>
<head>
<meta name="generator" content="pc-freak.net tidy">
<script src="https://ssl.google-analytics.com/urchin.js" type="text/javascript">
</script>
<script type="text/javascript">
_uacct = "UA-2102595-3";
urchinTracker();
</script>
<script type="text/javascript">
var gaJsHost = (("https:" == document.location.protocol) ? "https://ssl." : "http://");
document.write(unescape("%3Cscript src='" + gaJsHost + "google-analytics.com/ga.js' type='text/javascript'%3E%3C/script%3E"));
</script>
<script type="text/javascript">
try {
var pageTracker = _gat._getTracker("UA-2102595-6");
pageTracker._trackPageview();
} catch(err) {}
</script>

 

      –haproxy-protocol
              (HTTP) Send a HAProxy PROXY protocol v1 header at the beginning of the connection. This is used by some load balancers and reverse proxies
              to indicate the client's true IP address and port.

              This option is primarily useful when sending test requests to a service that expects this header.

              Added in 7.60.0.


7. Check apache log if remote Real Internet Source IPs are properly logged
 

root@webserver:~# tail -n 10 /var/log/apache2/access.log

213.183.53.58 – – [15/Apr/2022:22:18:59 +0300] "GET /proxy/browse.php?u=https%3A%2F%2Fsteamcommunity.com%2Fmarket%2Fitemordershistogram%3Fcountry HTTP/1.1" 200 12701 "https://www.pc-freak.net" "Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:98.0) Gecko/20100101 Firefox/98.0"
88.198.48.184 – – [15/Apr/2022:22:18:58 +0300] "GET /blog/iq-world-rank-country-smartest-nations/?cid=1330192 HTTP/1.1" 200 29574 "-" "Mozilla/5.0 (compatible; DataForSeoBot/1.0; +https://dataforseo.com/dataforseo-bot)"
213.183.53.58 – – [15/Apr/2022:22:19:00 +0300] "GET /proxy/browse.php?u=https%3A%2F%2Fsteamcommunity.com%2Fmarket%2Fitemordershistogram%3Fcountry
HTTP/1.1" 200 9080 "https://www.pc-freak.net" "Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:98.0) Gecko/20100101 Firefox/98.0"
159.223.65.16 – – [15/Apr/2022:22:19:01 +0300] "POST //blog//xmlrpc.php HTTP/1.1" 200 5477 "-" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/89.0.4389.114 Safari/537.36"
159.223.65.16 – – [15/Apr/2022:22:19:02 +0300] "POST //blog//xmlrpc.php HTTP/1.1" 200 5477 "-" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/89.0.4389.114 Safari/537.36"
213.91.190.233 – – [15/Apr/2022:22:19:02 +0300] "POST /blog/wp-admin/admin-ajax.php HTTP/1.1" 200 1243 "https://www.pc-freak.net/blog/wp-admin/post.php?post=16754&action=edit" "Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:89.0) Gecko/20100101 Firefox/89.0"
46.10.215.119 – – [15/Apr/2022:22:19:02 +0300] "GET /images/saint-Paul-and-Peter-holy-icon.jpg HTTP/1.1" 200 134501 "https://www.google.com/" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/100.0.4896.75 Safari/537.36 Edg/100.0.1185.39"
185.191.171.42 – – [15/Apr/2022:22:19:03 +0300] "GET /index.html.latest/tutorials/tutorials/penguins/vestnik/penguins/faith/vestnik/ HTTP/1.1" 200 11684 "-" "Mozilla/5.0 (compatible; SemrushBot/7~bl; +http://www.semrush.com/bot.html)"

116.179.37.243 – – [15/Apr/2022:22:19:50 +0300] "GET /blog/wp-content/cookieconsent.min.js HTTP/1.1" 200 7625 "https://www.pc-freak.net/blog/how-to-disable-nginx-static-requests-access-log-logging/" "Mozilla/5.0 (compatible; Baiduspider-render/2.0; +http://www.baidu.com/search/spider.html)"
116.179.37.237 – – [15/Apr/2022:22:19:50 +0300] "GET /blog/wp-content/plugins/google-analytics-dashboard-for-wp/assets/js/frontend-gtag.min.js?ver=7.5.0 HTTP/1.1" 200 8898 "https://www.pc-freak.net/blog/how-to-disable-nginx-static-requests-access-log-logging/" "Mozilla/5.0 (compatible; Baiduspider-render/2.0; +http://www.baidu.com/search/spider.html)"

 

You see from above output remote Source IPs in green are properly logged, so haproxy Cluster is correctly forwarding connections passing on in the Haproxy generated Initial header the Real IP of its remote connect IPs.


Sum it up, What was done?


HTTP_X_FORWARD_FOR is impossible to set, when haproxy is used on mode tcp and all traffic is sent as received from TCP IPv4 / IPv6 Network stack, e.g. modifying any HTTP sent traffic inside the headers is not possible as this might break up the data.

Thus Haproxy was configured to send all its received data by sending initial proxy header with the X_FORWARDED usual Source IP data, then remoteip Apache module was used to make Apache receive and understand haproxy sent Header which contains the original Source IP via the send-proxy functionality and example was given on how to test the remoteip on Webserver is working correctly.

Finally you've seen how to check configured haproxy and webserver are able to send and receive the End Client data with the originator real source IP correctly and those Internet IP is properly logged inside both haproxy and apaches.

Create Linux High Availability Load Balancer Cluster with Keepalived and Haproxy on Linux

Tuesday, March 15th, 2022

keepalived-logo-linux

Configuring a Linux HA (High Availibiltiy) for an Application with Haproxy is already used across many Websites on the Internet and serious corporations that has a crucial infrastructure has long time
adopted and used keepalived to provide High Availability Application level Clustering.
Usually companies choose to use HA Clusters with Haproxy with Pacemaker and Corosync cluster tools.
However one common used alternative solution if you don't have the oportunity to bring up a High availability cluster with Pacemaker / Corosync / pcs (Pacemaker Configuration System) due to fact machines you need to configure the cluster on are not Physical but VMWare Virtual Machines which couldn't not have configured a separate Admin Lans and Heartbeat Lan as we usually do on a Pacemaker Cluster due to the fact the 5 Ethernet LAN Card Interfaces of the VMWare Hypervisor hosts are configured as a BOND (e.g. all the incoming traffic to the VMWare vSphere  HV is received on one Virtual Bond interface).

I assume you have 2 separate vSphere Hypervisor Physical Machines in separate Racks and separate switches hosting the two VMs.
For the article, I'll call the two brand new brought Virtual Machines with some installation automation software such as Terraform or Ansible – vm-server1 and vm-server2 which would have configured some recent version of Linux.

In that scenario to have a High Avaiability for the VMs on Application level and assure at least one of the two is available at a time if one gets broken due toe malfunction of the HV, a Network connectivity issue, or because the VM OS has crashed.
Then one relatively easily solution is to use keepalived and configurea single High Availability Virtual IP (VIP) Address, i.e. 10.10.10.1, which would float among two VMs using keepalived so at a time at least one of the two VMs would be reachable on the Network.

haproxy_keepalived-vip-ip-diagram-linux

Having a VIP IP is quite a common solution in corporate world, as it makes it pretty easy to add F5 Load Balancer in front of the keepalived cluster setup to have a 3 Level of security isolation, which usually consists of:

1. Physical (access to the hardware or Virtualization hosts)
2. System Access (The mechanism to access the system login credetials users / passes, proxies, entry servers leading to DMZ-ed network)
3. Application Level (access to different programs behind L2 and data based on the specific identity of the individual user,
special Secondary UserID,  Factor authentication, biometrics etc.)

 

1. Install keepalived and haproxy on machines

Depending on the type of Linux OS:

On both machines
 

[root@server1:~]# yum install -y keepalived haproxy

If you have to install keepalived / haproxy on Debian / Ubuntu and other Deb based Linux distros

[root@server1:~]# apt install keepalived haproxy –yes

2. Configure haproxy (haproxy.cfg) on both server1 and server2

 

Create some /etc/haproxy/haproxy.cfg configuration

 

[root@server1:~]vim /etc/haproxy/haproxy.cfg

#———————————————————————
# Global settings
#———————————————————————
global
    log          127.0.0.1 local6 debug
    chroot       /var/lib/haproxy
    pidfile      /run/haproxy.pid
    stats socket /var/lib/haproxy/haproxy.sock mode 0600 level admin 
    maxconn      4000
    user         haproxy
    group        haproxy
    daemon
    #debug
    #quiet

#———————————————————————
# common defaults that all the 'listen' and 'backend' sections will
# use if not designated in their block
#———————————————————————
defaults
    mode        tcp
    log         global
#    option      dontlognull
#    option      httpclose
#    option      httplog
#    option      forwardfor
    option      redispatch
    option      log-health-checks
    timeout connect 10000 # default 10 second time out if a backend is not found
    timeout client 300000
    timeout server 300000
    maxconn     60000
    retries     3

#———————————————————————
# round robin balancing between the various backends
#———————————————————————

listen FRONTEND_APPNAME1
        bind 10.10.10.1:15000
        mode tcp
        option tcplog
#        #log global
        log-format [%t]\ %ci:%cp\ %bi:%bp\ %b/%s:%sp\ %Tw/%Tc/%Tt\ %B\ %ts\ %ac/%fc/%bc/%sc/%rc\ %sq/%bq
        balance roundrobin
        timeout client 350000
        timeout server 350000
        timeout connect 35000
        server app-server1 10.10.10.55:30000 weight 1 check port 68888
        server app-server2 10.10.10.55:30000 weight 2 check port 68888

listen FRONTEND_APPNAME2
        bind 10.10.10.1:15000
        mode tcp
        option tcplog
        #log global
        log-format [%t]\ %ci:%cp\ %bi:%bp\ %b/%s:%sp\ %Tw/%Tc/%Tt\ %B\ %ts\ %ac/%fc/%bc/%sc/%rc\ %sq/%bq
        balance roundrobin
        timeout client 350000
        timeout server 350000
        timeout connect 35000
        server app-server1 10.10.10.55:30000 weight 5
        server app-server2 10.10.10.55:30000 weight 5 

 

You can get a copy of above haproxy.cfg configuration here.
Once configured roll it on.

[root@server1:~]#  systemctl start haproxy
 
[root@server1:~]# ps -ef|grep -i hapro
root      285047       1  0 Mar07 ?        00:00:00 /usr/sbin/haproxy -Ws -f /etc/haproxy/haproxy.cfg -p /run/haproxy.pid
haproxy   285050  285047  0 Mar07 ?        00:00:26 /usr/sbin/haproxy -Ws -f /etc/haproxy/haproxy.cfg -p /run/haproxy.pid

Bring up the haproxy also on server2 machine, by placing same configuration and starting up the proxy.
 

[root@server1:~]vim /etc/haproxy/haproxy.cfg


 

3. Configure keepalived on both servers

We'll be configuring 2 nodes with keepalived even though if necessery this can be easily extended and you can add more nodes.
First we make a copy of the original or existing server configuration keepalived.conf (just in case we need it later on or if you already had something other configured manually by someone – that could be so on inherited servers by other sysadmin)
 

[root@server1:~]# mv /etc/keepalived/keepalived.conf /etc/keepalived/keepalived.conf.orig
[root@server2:~]# mv /etc/keepalived/keepalived.conf /etc/keepalived/keepalived.conf.orig

a. Configure keepalived to serve as a MASTER Node

 

[root@server1:~]# vim /etc/keepalived/keepalived.conf

Master Node
global_defs {
  router_id server1-fqdn # The hostname of this host.
  
  enable_script_security
  # Synchro of the state of the connections between the LBs on the eth0 interface
   lvs_sync_daemon eth0
 
notification_email {
        linuxadmin@notify-domain.com     # Email address for notifications 
    }
 notification_email_from keepalived@server1-fqdn        # The from address for the notifications
    smtp_server 127.0.0.1                       # SMTP server address
    smtp_connect_timeout 15
}

vrrp_script haproxy {
  script "killall -0 haproxy"
  interval 2
  weight 2
  user root
}

vrrp_instance LB_VIP_QA {
  virtual_router_id 50
  advert_int 1
  priority 51

  state MASTER
  interface eth0
  smtp_alert          # Enable Notifications Via Email
  
  authentication {
              auth_type PASS
              auth_pass testp141

    }
### Commented because running on VM on VMWare
##    unicast_src_ip 10.44.192.134 # Private IP address of master
##    unicast_peer {
##        10.44.192.135           # Private IP address of the backup haproxy
##   }

#        }
# master node with higher priority preferred node for Virtual IP if both keepalived up
###  priority 51
###  state MASTER
###  interface eth0
  virtual_ipaddress {
     10.10.10.1 dev eth0 # The virtual IP address that will be shared between MASTER and BACKUP
  }
  track_script {
      haproxy
  }
}

 

 To dowload a copy of the Master keepalived.conf configuration click here

Below are few interesting configuration variables, worthy to mention few words on, most of them are obvious by their names but for more clarity I'll also give a list here with short description of each:

 

  • vrrp_instance – defines an individual instance of the VRRP protocol running on an interface.
  • state – defines the initial state that the instance should start in (i.e. MASTER / SLAVE )state –
  • interface – defines the interface that VRRP runs on.
  • virtual_router_id – should be unique value per Keepalived Node (otherwise slave master won't function properly)
  • priority – the advertised priority, the higher the priority the more important the respective configured keepalived node is.
  • advert_int – specifies the frequency that advertisements are sent at (1 second, in this case).
  • authentication – specifies the information necessary for servers participating in VRRP to authenticate with each other. In this case, a simple password is defined.
    only the first eight (8) characters will be used as described in  to note is Important thing
    man keepalived.conf – keepalived.conf variables documentation !!! Nota Bene !!! – Password set on each node should match for nodes to be able to authenticate !
  • virtual_ipaddress – defines the IP addresses (there can be multiple) that VRRP is responsible for.
  • notification_email – the notification email to which Alerts will be send in case if keepalived on 1 node is stopped (e.g. the MASTER node switches from host 1 to 2)
  • notification_email_from – email address sender from where email will originte
    ! NB ! In order for notification_email to be working you need to have configured MTA or Mail Relay (set to local MTA) to another SMTP – e.g. have configured something like Postfix, Qmail or Postfix

b. Configure keepalived to serve as a SLAVE Node

[root@server1:~]vim /etc/keepalived/keepalived.conf
 

#Slave keepalived
global_defs {
  router_id server2-fqdn # The hostname of this host!

  enable_script_security
  # Synchro of the state of the connections between the LBs on the eth0 interface
  lvs_sync_daemon eth0
 
notification_email {
        linuxadmin@notify-host.com     # Email address for notifications
    }
 notification_email_from keepalived@server2-fqdn        # The from address for the notifications
    smtp_server 127.0.0.1                       # SMTP server address
    smtp_connect_timeout 15
}

vrrp_script haproxy {
  script "killall -0 haproxy"
  interval 2
  weight 2
  user root
}

vrrp_instance LB_VIP_QA {
  virtual_router_id 50
  advert_int 1
  priority 50

  state BACKUP
  interface eth0
  smtp_alert          # Enable Notifications Via Email

authentication {
              auth_type PASS
              auth_pass testp141
}
### Commented because running on VM on VMWare    
##    unicast_src_ip 10.10.192.135 # Private IP address of master
##    unicast_peer {
##        10.10.192.134         # Private IP address of the backup haproxy
##   }

###  priority 50
###  state BACKUP
###  interface eth0
  virtual_ipaddress {
     10.10.10.1 dev eth0 # The virtual IP address that will be shared betwee MASTER and BACKUP.
  }
  track_script {
    haproxy
  }
}

 

Download the keepalived.conf slave config here

 

c. Set required sysctl parameters for haproxy to work as expected
 

[root@server1:~]vim /etc/sysctl.conf
#Haproxy config
# haproxy
net.core.somaxconn=65535
net.ipv4.ip_local_port_range = 1024 65000
net.ipv4.ip_nonlocal_bind = 1
net.ipv4.tcp_tw_reuse = 1
net.ipv4.tcp_max_syn_backlog = 10240
net.ipv4.tcp_max_tw_buckets = 400000
net.ipv4.tcp_max_orphans = 60000
net.ipv4.tcp_synack_retries = 3

4. Test Keepalived keepalived.conf configuration syntax is OK

 

[root@server1:~]keepalived –config-test
(/etc/keepalived/keepalived.conf: Line 7) Unknown keyword 'lvs_sync_daemon_interface'
(/etc/keepalived/keepalived.conf: Line 21) Unable to set default user for vrrp script haproxy – removing
(/etc/keepalived/keepalived.conf: Line 31) (LB_VIP_QA) Specifying lvs_sync_daemon_interface against a vrrp is deprecated.
(/etc/keepalived/keepalived.conf: Line 31)              Please use global lvs_sync_daemon
(/etc/keepalived/keepalived.conf: Line 35) Truncating auth_pass to 8 characters
(/etc/keepalived/keepalived.conf: Line 50) (LB_VIP_QA) track script haproxy not found, ignoring…

I've experienced this error because first time I've configured keepalived, I did not mention the user with which the vrrp script haproxy should run,
in prior versions of keepalived, leaving the field empty did automatically assumed you have the user with which the vrrp script runs to be set to root
as of RHELs keepalived-2.1.5-6.el8.x86_64, i've been using however this is no longer so and thus in prior configuration as you can see I've
set the user in respective section to root.
The error Unknown keyword 'lvs_sync_daemon_interface'
is also easily fixable by just substituting the lvs_sync_daemon_interface and lvs_sync_daemon and reloading
keepalived etc.

Once keepalived is started and you can see the process on both machines running in process list.

[root@server1:~]ps -ef |grep -i keepalived
root     1190884       1  0 18:50 ?        00:00:00 /usr/sbin/keepalived -D
root     1190885 1190884  0 18:50 ?        00:00:00 /usr/sbin/keepalived -D

Next step is to check the keepalived statuses as well as /var/log/keepalived.log

If everything is configured as expected on both keepalived on first node you should see one is master and one is slave either in the status or the log

[root@server1:~]#systemctl restart keepalived

 

[root@server1:~]systemctl status keepalived|grep -i state
Mar 14 18:59:02 server1-fqdn Keepalived_vrrp[1192003]: (LB_VIP_QA) Entering MASTER STATE

[root@server1:~]systemctl status keepalived

● keepalived.service – LVS and VRRP High Availability Monitor
   Loaded: loaded (/usr/lib/systemd/system/keepalived.service; enabled; vendor preset: disabled)
   Active: inactive (dead) since Mon 2022-03-14 18:15:51 CET; 32min ago
  Process: 1187587 ExecStart=/usr/sbin/keepalived $KEEPALIVED_OPTIONS (code=exited, status=0/SUCCESS)
 Main PID: 1187589 (code=exited, status=0/SUCCESS)

Mar 14 18:15:04 server1lb-fqdn Keepalived_vrrp[1187590]: Sending gratuitous ARP on eth0 for 10.44.192.142
Mar 14 18:15:50 server1lb-fqdn systemd[1]: Stopping LVS and VRRP High Availability Monitor…
Mar 14 18:15:50 server1lb-fqdn Keepalived[1187589]: Stopping
Mar 14 18:15:50 server1lb-fqdn Keepalived_vrrp[1187590]: (LB_VIP_QA) sent 0 priority
Mar 14 18:15:50 server1lb-fqdn Keepalived_vrrp[1187590]: (LB_VIP_QA) removing VIPs.
Mar 14 18:15:51 server1lb-fqdn Keepalived_vrrp[1187590]: Stopped – used 0.002007 user time, 0.016303 system time
Mar 14 18:15:51 server1lb-fqdn Keepalived[1187589]: CPU usage (self/children) user: 0.000000/0.038715 system: 0.001061/0.166434
Mar 14 18:15:51 server1lb-fqdn Keepalived[1187589]: Stopped Keepalived v2.1.5 (07/13,2020)
Mar 14 18:15:51 server1lb-fqdn systemd[1]: keepalived.service: Succeeded.
Mar 14 18:15:51 server1lb-fqdn systemd[1]: Stopped LVS and VRRP High Availability Monitor

[root@server2:~]systemctl status keepalived|grep -i state
Mar 14 18:59:02 server2-fqdn Keepalived_vrrp[297368]: (LB_VIP_QA) Entering BACKUP STATE

[root@server1:~]# grep -i state /var/log/keepalived.log
Mar 14 18:59:02 server1lb-fqdn Keepalived_vrrp[297368]: (LB_VIP_QA) Entering MASTER STATE
 

a. Fix Keepalived SECURITY VIOLATION – scripts are being executed but script_security not enabled.
 

When configurating keepalived for a first time we have faced the following strange error inside keepalived status inside keepalived.log 
 

Feb 23 14:28:41 server1 Keepalived_vrrp[945478]: SECURITY VIOLATION – scripts are being executed but script_security not enabled.

 

To fix keepalived SECURITY VIOLATION error:

Add to /etc/keepalived/keepalived.conf on the keepalived node hosts
inside 

global_defs {}

After chunk
 

enable_script_security

include

# Synchro of the state of the connections between the LBs on the eth0 interface
  lvs_sync_daemon_interface eth0

 

5. Prepare rsyslog configuration and Inlcude additional keepalived options
to force keepalived log into /var/log/keepalived.log

To force keepalived log into /var/log/keepalived.log on RHEL 8 / CentOS and other Redhat Package Manager (RPM) Linux distributions

[root@server1:~]# vim /etc/rsyslog.d/48_keepalived.conf

#2022/02/02: HAProxy logs to local6, save the messages
local7.*                                                /var/log/keepalived.log
if ($programname == 'Keepalived') then -/var/log/keepalived.log
if ($programname == 'Keepalived_vrrp') then -/var/log/keepalived.log
& stop

[root@server:~]# touch /var/log/keepalived.log

Reload rsyslog to load new config
 

[root@server:~]# systemctl restart rsyslog
[root@server:~]# systemctl status rsyslog

 

rsyslog.service – System Logging Service
   Loaded: loaded (/usr/lib/systemd/system/rsyslog.service; enabled; vendor preset: enabled)
  Drop-In: /etc/systemd/system/rsyslog.service.d
           └─rsyslog-service.conf
   Active: active (running) since Mon 2022-03-07 13:34:38 CET; 1 weeks 0 days ago
     Docs: man:rsyslogd(8)

           https://www.rsyslog.com/doc/
 Main PID: 269574 (rsyslogd)
    Tasks: 6 (limit: 100914)
   Memory: 5.1M
   CGroup: /system.slice/rsyslog.service
           └─269574 /usr/sbin/rsyslogd -n

Mar 15 08:15:16 server1lb-fqdn rsyslogd[269574]: — MARK —
Mar 15 08:35:16 server1lb-fqdn rsyslogd[269574]: — MARK —
Mar 15 08:55:16 server1lb-fqdn rsyslogd[269574]: — MARK —

 

If once keepalived is loaded but you still have no log written inside /var/log/keepalived.log

[root@server1:~]# vim /etc/sysconfig/keepalived
 KEEPALIVED_OPTIONS="-D -S 7"

[root@server2:~]# vim /etc/sysconfig/keepalived
 KEEPALIVED_OPTIONS="-D -S 7"

[root@server1:~]# systemctl restart keepalived.service
[root@server1:~]#  systemctl status keepalived

● keepalived.service – LVS and VRRP High Availability Monitor
   Loaded: loaded (/usr/lib/systemd/system/keepalived.service; enabled; vendor preset: disabled)
   Active: active (running) since Thu 2022-02-24 12:12:20 CET; 2 weeks 4 days ago
 Main PID: 1030501 (keepalived)
    Tasks: 2 (limit: 100914)
   Memory: 1.8M
   CGroup: /system.slice/keepalived.service
           ├─1030501 /usr/sbin/keepalived -D
           └─1030502 /usr/sbin/keepalived -D

Warning: Journal has been rotated since unit was started. Log output is incomplete or unavailable.

[root@server2:~]# systemctl restart keepalived.service
[root@server2:~]# systemctl status keepalived

6. Monitoring VRRP traffic of the two keepaliveds with tcpdump
 

Once both keepalived are up and running a good thing is to check the VRRP protocol traffic keeps fluently on both machines.
Keepalived VRRP keeps communicating over the TCP / IP Port 112 thus you can simply snoop TCP tracffic on its protocol.
 

[root@server1:~]# tcpdump proto 112

tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), capture size 262144 bytes
11:08:07.356187 IP server1lb-fqdn > vrrp.mcast.net: VRRPv2, Advertisement, vrid 50, prio 53, authtype simple, intvl 1s, length 20
11:08:08.356297 IP server1lb-fqdn > vrrp.mcast.net: VRRPv2, Advertisement, vrid 50, prio 53, authtype simple, intvl 1s, length 20
11:08:09.356408 IP server1lb-fqdn > vrrp.mcast.net: VRRPv2, Advertisement, vrid 50, prio 53, authtype simple, intvl 1s, length 20
11:08:10.356511 IP server1lb-fqdn > vrrp.mcast.net: VRRPv2, Advertisement, vrid 50, prio 53, authtype simple, intvl 1s, length 20
11:08:11.356655 IP server1lb-fqdn > vrrp.mcast.net: VRRPv2, Advertisement, vrid 50, prio 53, authtype simple, intvl 1s, length 20

[root@server2:~]# tcpdump proto 112

tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
​listening on eth0, link-type EN10MB (Ethernet), capture size 262144 bytes
11:08:07.356187 IP server1lb-fqdn > vrrp.mcast.net: VRRPv2, Advertisement, vrid 50, prio 53, authtype simple, intvl 1s, length 20
11:08:08.356297 IP server1lb-fqdn > vrrp.mcast.net: VRRPv2, Advertisement, vrid 50, prio 53, authtype simple, intvl 1s, length 20
11:08:09.356408 IP server1lb-fqdn > vrrp.mcast.net: VRRPv2, Advertisement, vrid 50, prio 53, authtype simple, intvl 1s, length 20
11:08:10.356511 IP server1lb-fqdn > vrrp.mcast.net: VRRPv2, Advertisement, vrid 50, prio 53, authtype simple, intvl 1s, length 20
11:08:11.356655 IP server1lb-fqdn > vrrp.mcast.net: VRRPv2, Advertisement, vrid 50, prio 53, authtype simple, intvl 1s, length 20

As you can see the VRRP traffic on the network is originating only from server1lb-fqdn, this is so because host server1lb-fqdn is the keepalived configured master node.

It is possible to spoof the password configured to authenticate between two nodes, thus if you're bringing up keepalived service cluster make sure your security is tight at best the machines should be in a special local LAN DMZ, do not configure DMZ on the internet !!! 🙂 Or if you eventually decide to configure keepalived in between remote hosts, make sure you somehow use encrypted VPN or SSH tunnels to tunnel the VRRP traffic.

[root@server1:~]tcpdump proto 112 -vv
tcpdump: listening on eth0, link-type EN10MB (Ethernet), capture size 262144 bytes
11:36:25.530772 IP (tos 0xc0, ttl 255, id 59838, offset 0, flags [none], proto VRRP (112), length 40)
    server1lb-fqdn > vrrp.mcast.net: vrrp server1lb-fqdn > vrrp.mcast.net: VRRPv2, Advertisement, vrid 50, prio 53, authtype simple, intvl 1s, length 20, addrs: VIPIP_QA auth "testp431"
11:36:26.530874 IP (tos 0xc0, ttl 255, id 59839, offset 0, flags [none], proto VRRP (112), length 40)
    server1lb-fqdn > vrrp.mcast.net: vrrp server1lb-fqdn > vrrp.mcast.net: VRRPv2, Advertisement, vrid 50, prio 53, authtype simple, intvl 1s, length 20, addrs: VIPIP_QA auth "testp431"

Lets also check what floating IP is configured on the machines:

[root@server1:~]# ip -brief address show
lo               UNKNOWN        127.0.0.1/8 
eth0             UP             10.10.10.5/26 10.10.10.1/32 

The 10.10.10.5 IP is the main IP set on LAN interface eth0, 10.10.10.1 is the floating IP which as you can see is currently set by keepalived to listen on first node.

[root@server2:~]# ip -brief address show |grep -i 10.10.10.1

An empty output is returned as floating IP is currently configured on server1

To double assure ourselves the IP is assigned on correct machine, lets ping it and check the IP assigned MAC  currently belongs to which machine.
 

[root@server2:~]# ping 10.10.10.1
PING 10.10.10.1 (10.10.10.1) 56(84) bytes of data.
64 bytes from 10.10.10.1: icmp_seq=1 ttl=64 time=0.526 ms
^C
— 10.10.10.1 ping statistics —
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 0.526/0.526/0.526/0.000 ms

[root@server2:~]# arp -an |grep -i 10.44.192.142
? (10.10.10.1) at 00:48:54:91:83:7d [ether] on eth0
[root@server2:~]# ip a s|grep -i 00:48:54:91:83:7d
[root@server2:~]# 

As you can see from below output MAC is not found in configured IPs on server2.
 

[root@server1-fqdn:~]# /sbin/ip a s|grep -i 00:48:54:91:83:7d -B1 -A1
 eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
    link/ether 00:48:54:91:83:7d brd ff:ff:ff:ff:ff:ff
inet 10.10.10.1/26 brd 10.10.1.191 scope global noprefixroute eth0

Pretty much expected MAC is on keepalived node server1.

 

7. Testing keepalived on server1 and server2 maachines VIP floating IP really works
 

To test the overall configuration just created, you should stop keeaplived on the Master node and in meantime keep an eye on Slave node (server2), whether it can figure out the Master node is gone and switch its
state BACKUP to save MASTER. By changing the secondary (Slave) keepalived to master the floating IP: 10.10.10.1 will be brought up by the scripts on server2.

Lets assume that something went wrong with server1 VM host, for example the machine crashed due to service overload, DDoS or simply a kernel bug or whatever reason.
To simulate that we simply have to stop keepalived, then the broadcasted information on VRRP TCP/IP proto port 112 will be no longer available and keepalived on node server2, once
unable to communicate to server1 should chnage itself to state MASTER.

[root@server1:~]# systemctl stop keepalived
[root@server1:~]# systemctl status keepalived

● keepalived.service – LVS and VRRP High Availability Monitor
   Loaded: loaded (/usr/lib/systemd/system/keepalived.service; enabled; vendor preset: disabled)
   Active: inactive (dead) since Tue 2022-03-15 12:11:33 CET; 3s ago
  Process: 1192001 ExecStart=/usr/sbin/keepalived $KEEPALIVED_OPTIONS (code=exited, status=0/SUCCESS)
 Main PID: 1192002 (code=exited, status=0/SUCCESS)

Mar 14 18:59:07 server1lb-fqdn Keepalived_vrrp[1192003]: Sending gratuitous ARP on eth0 for 10.10.10.1
Mar 15 12:11:32 server1lb-fqdn systemd[1]: Stopping LVS and VRRP High Availability Monitor…
Mar 15 12:11:32 server1lb-fqdn Keepalived[1192002]: Stopping
Mar 15 12:11:32 server1lb-fqdn Keepalived_vrrp[1192003]: (LB_VIP_QA) sent 0 priority
Mar 15 12:11:32 server1lb-fqdn Keepalived_vrrp[1192003]: (LB_VIP_QA) removing VIPs.
Mar 15 12:11:33 server1lb-fqdn Keepalived_vrrp[1192003]: Stopped – used 2.145252 user time, 15.513454 system time
Mar 15 12:11:33 server1lb-fqdn Keepalived[1192002]: CPU usage (self/children) user: 0.000000/44.555362 system: 0.001151/170.118126
Mar 15 12:11:33 server1lb-fqdn Keepalived[1192002]: Stopped Keepalived v2.1.5 (07/13,2020)
Mar 15 12:11:33 server1lb-fqdn systemd[1]: keepalived.service: Succeeded.
Mar 15 12:11:33 server1lb-fqdn systemd[1]: Stopped LVS and VRRP High Availability Monitor.

 

On keepalived off, you will get also a notification Email on the Receipt Email configured from keepalived.conf from the working keepalived node with a simple message like:

=> VRRP Instance is no longer owning VRRP VIPs <=

Once keepalived is back up you will get another notification like:

=> VRRP Instance is now owning VRRP VIPs <=

[root@server2:~]# systemctl status keepalived
● keepalived.service – LVS and VRRP High Availability Monitor
   Loaded: loaded (/usr/lib/systemd/system/keepalived.service; enabled; vendor preset: disabled)
   Active: active (running) since Mon 2022-03-14 18:13:52 CET; 17h ago
  Process: 297366 ExecStart=/usr/sbin/keepalived $KEEPALIVED_OPTIONS (code=exited, status=0/SUCCESS)
 Main PID: 297367 (keepalived)
    Tasks: 2 (limit: 100914)
   Memory: 2.1M
   CGroup: /system.slice/keepalived.service
           ├─297367 /usr/sbin/keepalived -D -S 7
           └─297368 /usr/sbin/keepalived -D -S 7

Mar 15 12:11:33 server2lb-fqdn Keepalived_vrrp[297368]: Sending gratuitous ARP on eth0 for 10.10.10.1
Mar 15 12:11:33 server2lb-fqdn Keepalived_vrrp[297368]: Sending gratuitous ARP on eth0 for 10.10.10.1
Mar 15 12:11:33 server2lb-fqdn Keepalived_vrrp[297368]: Remote SMTP server [127.0.0.1]:25 connected.
Mar 15 12:11:33 server2lb-fqdn Keepalived_vrrp[297368]: SMTP alert successfully sent.
Mar 15 12:11:38 server2lb-fqdn Keepalived_vrrp[297368]: (LB_VIP_QA) Sending/queueing gratuitous ARPs on eth0 for 10.10.10.1
Mar 15 12:11:38 server2lb-fqdn Keepalived_vrrp[297368]: Sending gratuitous ARP on eth0 for 10.10.10.1
Mar 15 12:11:38 server2lb-fqdn Keepalived_vrrp[297368]: Sending gratuitous ARP on eth0 for 10.10.10.1
Mar 15 12:11:38 server2lb-fqdn Keepalived_vrrp[297368]: Sending gratuitous ARP on eth0 for 10.10.10.1
Mar 15 12:11:38 server2lb-fqdn Keepalived_vrrp[297368]: Sending gratuitous ARP on eth0 for 10.10.10.1
Mar 15 12:11:38 server2lb-fqdn Keepalived_vrrp[297368]: Sending gratuitous ARP on eth0 for 10.10.10.1

[root@server2:~]#  ip addr show|grep -i 10.10.10.1
    inet 10.10.10.1/32 scope global eth0
    

As you see the VIP is now set on server2, just like expected – that's OK, everything works as expected. If the IP did not move double check the keepalived.conf on both nodes for errors or misconfigurations.

To recover the initial order of things so server1 is MASTER and server2 SLAVE host, we just have to switch on the keepalived on server1 machine.

[root@server1:~]# systemctl start keepalived

The automatic change of server1 to MASTER node and respective move of the VIP IP is done because of the higher priority (of importance we previously configured on server1 in keepalived.conf).
 

What we learned?
 

So what we learned in  this article?
We have seen how to easily install and configure a High Availability Load balancer with Keepalived with single floating VIP IP address with 1 MASTER and 1 SLAVE host and a Haproxy example config with few frontends / App backends. We have seen how the config can be tested for potential errors and how we can monitor whether the VRRP2 network traffic flows between nodes and how to potentially debug it further if necessery.
Further on rawly explained some of the keepalived configurations but as keepalived can do pretty much more,for anyone seriously willing to deal with keepalived on a daily basis or just fine tune some already existing ones, you better read closely its manual page "man keepalived.conf" as well as the official Redhat Linux documentation page on setting up a Linux cluster with Keepalived (Be prepare for a small nightmare as the documentation of it seems to be a bit chaotic, and even I would say partly missing or opening questions on what does the developers did meant – not strange considering the havoc that is pretty much as everywhere these days.)

Finally once keepalived hosts are prepared, it was shown how to test the keepalived application cluster and Floating IP does move between nodes in case if one of the 2 keepalived nodes is inaccessible.

The same logic can be repeated multiple times and if necessery you can set multiple VIPs to expand the HA reachable IPs solution.

high-availability-with-two-vips-example-diagram

The presented idea is with haproxy forward Proxy server to proxy requests towards Application backend (servince machines), however if you need to set another set of server on the flow to  process HTML / XHTML / PHP / Perl / Python  programming code, with some common Webserver setup ( Nginx / Apache / Tomcat / JBOSS) and enable SSL Secure certificate with lets say Letsencrypt, this can be relatively easily done. If you want to implement letsencrypt and a webserver check this redundant SSL Load Balancing with haproxy & keepalived article.

That's all folks, hope you enjoyed.
If you need to configure keepalived Cluster or a consultancy write your query here 🙂

Fix weird double logging in haproxy.log file due to haproxy.cfg misconfiguration

Tuesday, March 8th, 2022

haproxy-logging-front-network-back-network-diagram

While we were building a new machine that will serve as a Haproxy server proxy frontend to tunnel some traffic to a number of backends, 
came across weird oddity. The call requests sent to the haproxy and redirected to backend servers, were being written in the log twice.

Since we have two backend Application servers hat are serving the request, my first guess was this is caused by the fact haproxy
tries to connect to both nodes on each sent request, or that the double logging is caused by the rsyslogd doing something strange on each
received query. The rsyslog configuration configured to send via local6 facility to rsyslog is like that:

$ModLoad imudp
$UDPServerAddress 127.0.0.1
$UDPServerRun 514
#2022/02/02: HAProxy logs to local6, save the messages
local6.*                                                /var/log/haproxy.log


The haproxy basic global and defaults and frontend section config is like that:
 

global
    log          127.0.0.1 local6 debug
    chroot       /var/lib/haproxy
    pidfile      /run/haproxy.pid
    stats socket /var/lib/haproxy/haproxy.sock mode 0600 level admin
    maxconn      4000
    user         haproxy
    group        haproxy
    daemon

 

defaults
    mode        tcp
    log         global
#    option      dontlognull
#    option      httpclose
#    option      httplog
#    option      forwardfor
    option      redispatch
    option      log-health-checks
    timeout connect 10000 # default 10 second time out if a backend is not found
    timeout client 300000
    timeout server 300000
    maxconn     60000
    retries     3
 

listen FRONTEND1
        bind 10.71.80.5:63750
        mode tcp
        option tcplog
        log global
        log-format [%t]\ %ci:%cp\ %bi:%bp\ %b/%s:%sp\ %Tw/%Tc/%Tt\ %B\ %ts\ %ac/%fc/%bc/%sc/%rc\ %sq/%bq
        balance roundrobin
        timeout client 350000
        timeout server 350000
        timeout connect 35000
        server backend-host1 10.80.50.1:13750 weight 1 check port 15000
        server backend-host2 10.80.50.2:13750 weight 2 check port 15000

After quick research online on why this odd double logging of request ends in /var/log/haproxy.log it turns out this is caused by the 

log global  defined double under the defaults section as well as in the frontend itself, hence to resolve simply had to comment out the log global in Frontend, so it looks like so:

listen FRONTEND1
        bind 10.71.80.5:63750
        mode tcp
        option tcplog
  #      log global
        log-format [%t]\ %ci:%cp\ %bi:%bp\ %b/%s:%sp\ %Tw/%Tc/%Tt\ %B\ %ts\ %ac/%fc/%bc/%sc/%rc\ %sq/%bq
        balance roundrobin
        timeout client 350000
        timeout server 350000
        timeout connect 35000
        server backend-host1 10.80.50.1:13750 weight 1 check port 15000
        server backend-host2 10.80.50.2:13750 weight 2 check port 15000

 


Next just reloaded haproxy and one request started leaving only one trail inside haproxy.log as expected 🙂
 

How to configure haproxy logging to separate file on Redhat Enterprise Linux 8.5 Ootpa

Thursday, February 3rd, 2022

haproxy-rsyslog-architecture-logging-picture

Configuring proper logging for haproxy is always a pain in the ass in Linux, because of rsyslogd various config syntax among versions, because of bugs in OS etc. 
Today we have been given 2 Redhat 8.5 Linux servers where we had a task to start configuring haproxies, to have an idea on what is going on of course we had to enable proper haproxy logging in separate log file under separate local, for the test one can use haproxy's 

log /dev/log local6

config, this is a general way to configure logging which I've described earlier in the article How to enable haproxy logging to a separate log /var/log/haproxy.log / prevent duplicate messages to appear in /var/log/messages
However this time I wanted to not use /dev/log as this device is also used by systemd / journald and theoretically could be used by other services and there might be multiple services logging to the same places possibly leading to some issue, thus I wanted to send and process the haproxy messages directly from rsyslog on RHEL 8.5.

Create a custom file that is loaded with the rest of configuration from /etc/rsyslog.conf with a line like:
 

# Include all config files in /etc/rsyslog.d/
include(file="/etc/rsyslog.d/*.conf" mode="optional")


Create 49_haproxy.conf with below content

[root@haproxy: ~]# vim /etc/rsyslog.d/49_haproxy.conf

$ModLoad imudp
$UDPServerAddress 127.0.0.1
$UDPServerRun 514
#2022/02/02: HAProxy logs to local6, save the messages
local6.*                                                /var/log/haproxy.log
if ($programname == 'haproxy') then -/var/log/haproxy.log
& stop

touch /var/log/haproxy.log
chown haproxy:haproxy /var/log/haproxy.log

In /etc/haproxy/haproxy.cfg under global section to print in verbose mode messages (i.e. check, the haproxy is receiving properly sent traffic) do configure something like:

 

global
  log          127.0.0.1 local6 debug


Eventually you might want to remove the debug word out of the config, if you don't want to log too much verbosily once everything is properly tested and configured

[root@haproxy: ~]# curl -v -c -k 10.10.192.135:16010
* Rebuilt URL to: 10.10.192.135:15010/
*   Trying 10.10.192.135…
* TCP_NODELAY set
* Connected to 10.10.192.135 (10.10.192.135) port 15010 (#0)
> GET / HTTP/1.1
> Host: 10.10.192.135:15010
> User-Agent: curl/7.61.1
> Accept: */*

* Empty reply from server
* Connection #0 to host 10.10.192.135 left intact
curl: (52) Empty reply from server

 

In /var/log/haproxy.log you should get some messages like:
 

Feb  3 14:16:44 localhost.localdomain haproxy[25029]: proxy IN_Traffic_Bak has no server available!
Feb  3 14:16:44 localhost.localdomain haproxy[25029]: proxy IN_Traffic_Bak has no server available!
Feb  3 15:59:50 localhost.localdomain haproxy[25029]: [03/Feb/2022:15:59:50.162] 10.44.192.135:1348 -:- IN_Traffic/<NOSRV>:- -1/-1/0 0 SC 1/1/0/0/0 0/0
Feb  3 15:59:50 localhost.localdomain haproxy[25029]: [03/Feb/2022:15:59:50.162] 10.44.192.135:1348 -:- IN_Traffic/<NOSRV>:- -1/-1/0 0 SC 1/1/0/0/0 0/0
Feb  3 15:59:50 localhost.localdomain haproxy[25029]: [03/Feb/2022:15:59:50.162] 10.44.192.135:1348 -:- IN_Traffic/<NOSRV>:- -1/-1/0 0 SC 1/1/0/0/0 0/0

 

How to enable HaProxy logging to a separate log /var/log/haproxy.log / prevent HAProxy duplicate messages to appear in /var/log/messages

Wednesday, February 19th, 2020

haproxy-logging-basics-how-to-log-to-separate-file-prevent-duplicate-messages-haproxy-haproxy-weblogo-squares
haproxy  logging can be managed in different form the most straight forward way is to directly use /dev/log either you can configure it to use some log management service as syslog or rsyslogd for that.

If you don't use rsyslog yet to install it: 

# apt install -y rsyslog

Then to activate logging via rsyslogd we can should add either to /etc/rsyslogd.conf or create a separte file and include it via /etc/rsyslogd.conf with following content:
 

Enable haproxy logging from rsyslogd


Log haproxy messages to separate log file you can use some of the usual syslog local0 to local7 locally used descriptors inside the conf (be aware that if you try to use some wrong value like local8, local9 as a logging facility you will get with empty haproxy.log, even though the permissions of /var/log/haproxy.log are readable and owned by haproxy user.

When logging to a local Syslog service, writing to a UNIX socket can be faster than targeting the TCP loopback address. Generally, on Linux systems, a UNIX socket listening for Syslog messages is available at /dev/log because this is where the syslog() function of the GNU C library is sending messages by default. To address UNIX socket in haproxy.cfg use:

log /dev/log local2 


If you want to log into separate log each of multiple running haproxy instances with different haproxy*.cfg add to /etc/rsyslog.conf lines like:

local2.* -/var/log/haproxylog2.log
local3.* -/var/log/haproxylog3.log


One important note to make here is since rsyslogd is used for haproxy logging you need to have enabled in rsyslogd imudp and have a UDP port listener on the machine.

E.g. somewhere in rsyslog.conf or via rsyslog include file from /etc/rsyslog.d/*.conf needs to have defined following lines:

$ModLoad imudp
$UDPServerRun 514


I prefer to use external /etc/rsyslog.d/20-haproxy.conf include file that is loaded and enabled rsyslogd via /etc/rsyslog.conf:

# vim /etc/rsyslog.d/20-haproxy.conf

$ModLoad imudp
$UDPServerRun 514​
local2.* -/var/log/haproxy2.log


It is also possible to produce different haproxy log output based on the severiy to differentiate between important and less important messages, to do so you'll need to rsyslog.conf something like:
 

# Creating separate log files based on the severity
local0.* /var/log/haproxy-traffic.log
local0.notice /var/log/haproxy-admin.log

 

Prevent Haproxy duplicate messages to appear in /var/log/messages

If you use local2 and some default rsyslog configuration then you will end up with the messages coming from haproxy towards local2 facility producing doubled simultaneous records to both your pre-defined /var/log/haproxy.log and /var/log/messages on Proxy servers that receive few thousands of simultanous connections per second.
This is a problem since doubling the log will produce too much data and on systems with smaller /var/ partition you will quickly run out of space + this haproxy requests logging to /var/log/messages makes the file quite unreadable for normal system events which are so important to track clearly what is happening on the server daily.

To prevent the haproxy duplicate messages you need to define somewhere in rsyslogd usually /etc/rsyslog.conf local2.none near line of facilities configured to log to file:

*.info;mail.none;authpriv.none;cron.none;local2.none     /var/log/messages

This configuration should work but is more rarely used as most people prefer to have haproxy log being written not directly to /dev/log which is used by other services such as syslogd / rsyslogd.

To use /dev/log to output logs from haproxy configuration in global section use config like:
 

global
        log /dev/log local2 debug
        chroot /var/lib/haproxy
        stats socket /run/haproxy/admin.sock mode 660 level admin
        stats timeout 30s
        user haproxy
        group haproxy
        daemon

The log global directive basically says, use the log line that was set in the global section for whole config till end of file. Putting a log global directive into the defaults section is equivalent to putting it into all of the subsequent proxy sections.

Using global logging rules is the most common HAProxy setup, but you can put them directly into a frontend section instead. It can be useful to have a different logging configuration as a one-off. For example, you might want to point to a different target Syslog server, use a different logging facility, or capture different severity levels depending on the use case of the backend application. 

Insetad of using /dev/log interface that is on many distributions heavily used by systemd to store / manage and distribute logs,  many haproxy server sysadmins nowdays prefer to use rsyslogd as a default logging facility that will manage haproxy logs.
Admins prefer to use some kind of mediator service to manage log writting such as rsyslogd or syslog, the reason behind might vary but perhaps most important reason is  by using rsyslogd it is possible to write logs simultaneously locally on disk and also forward logs  to a remote Logging server  running rsyslogd service.

Logging is defined in /etc/haproxy/haproxy.cfg or the respective configuration through global section but could be also configured to do a separate logging based on each of the defined Frontend Backends or default section. 
A sample exceprt from this section looks something like:

#———————————————————————
# Global settings
#———————————————————————
global
    log         127.0.0.1 local2

    chroot      /var/lib/haproxy
    pidfile     /var/run/haproxy.pid
    maxconn     4000
    user        haproxy
    group       haproxy
    daemon

    # turn on stats unix socket
    stats socket /var/lib/haproxy/stats

#———————————————————————
defaults
    mode                    tcp
    log                     global
    option                  tcplog
    #option                  dontlognull
    #option http-server-close
    #option forwardfor       except 127.0.0.0/8
    option                  redispatch
    retries                 7
    #timeout http-request    10s
    timeout queue           10m
    timeout connect         30s
    timeout client          20m
    timeout server          10m
    #timeout http-keep-alive 10s
    timeout check           30s
    maxconn                 3000

 

 

# HAProxy Monitoring Config
#———————————————————————
listen stats 192.168.0.5:8080                #Haproxy Monitoring run on port 8080
    mode http
    option httplog
    option http-server-close
    stats enable
    stats show-legends
    stats refresh 5s
    stats uri /stats                            #URL for HAProxy monitoring
    stats realm Haproxy\ Statistics
    stats auth hproxyauser:Password___          #User and Password for login to the monitoring dashboard

 

#———————————————————————
# frontend which proxys to the backends
#———————————————————————
frontend ft_DKV_PROD_WLPFO
    mode tcp
    bind 192.168.233.5:30000-31050
    option tcplog
    log-format %ci:%cp\ [%t]\ %ft\ %b/%s\ %Tw/%Tc/%Tt\ %B\ %ts\ %ac/%fc/%bc/%sc/%rc\ %sq/%bq
    default_backend Default_Bakend_Name


#———————————————————————
# round robin balancing between the various backends
#———————————————————————
backend bk_DKV_PROD_WLPFO
    mode tcp
    # (0) Load Balancing Method.
    balance source
    # (4) Peer Sync: a sticky session is a session maintained by persistence
    stick-table type ip size 1m peers hapeers expire 60m
    stick on src
    # (5) Server List
    # (5.1) Backend
    server Backend_Server1 10.10.10.1 check port 18088
    server Backend_Server2 10.10.10.2 check port 18088 backup


The log directive in above config instructs HAProxy to send logs to the Syslog server listening at 127.0.0.1:514. Messages are sent with facility local2, which is one of the standard, user-defined Syslog facilities. It’s also the facility that our rsyslog configuration is expecting. You can add more than one log statement to send output to multiple Syslog servers.

Once rsyslog and haproxy logging is configured as a minumum you need to restart rsyslog (assuming that haproxy config is already properly loaded):

# systemctl restart rsyslogd.service

To make sure rsyslog reloaded successfully:

systemctl status rsyslogd.service


Restarting HAproxy

If the rsyslogd logging to 127.0.0.1 port 514 was recently added a HAProxy restart should also be run, you can do it with:
 

# /usr/sbin/haproxy -f /etc/haproxy/haproxy.cfg -D -p /var/run/haproxy.pid -sf $(cat /var/run/haproxy.pid)


Or to restart use systemctl script (if haproxy is not used in a cluster with corosync / heartbeat).

# systemctl restart haproxy.service

You can control how much information is logged by adding a Syslog level by

    log         127.0.0.1 local2 info


The accepted values are the standard syslog security level severity:

Value Severity Keyword Deprecated keywords Description Condition
0 Emergency emerg panic System is unusable A panic condition.
1 Alert alert   Action must be taken immediately A condition that should be corrected immediately, such as a corrupted system database.
2 Critical crit   Critical conditions Hard device errors.
3 Error err error Error conditions  
4 Warning warning warn Warning conditions  
5 Notice notice   Normal but significant conditions Conditions that are not error conditions, but that may require special handling.
6 Informational info   Informational messages  
7 Debug debug   Debug-level messages Messages that contain information normally of use only when debugging a program.

 

Logging only errors / timeouts / retries and errors is done with option:

Note that if the rsyslog is configured to listen on different port for some weird reason you should not forget to set the proper listen port, e.g.:
 

  log         127.0.0.1:514 local2 info

option dontlog-normal

in defaults or frontend section.

You most likely want to enable this only during certain times, such as when performing benchmarking tests.

(or log-format-sd for structured-data syslog) directive in your defaults or frontend
 

Haproxy Logging shortly explained


The type of logging you’ll see is determined by the proxy mode that you set within HAProxy. HAProxy can operate either as a Layer 4 (TCP) proxy or as Layer 7 (HTTP) proxy. TCP mode is the default. In this mode, a full-duplex connection is established between clients and servers, and no layer 7 examination will be performed. When in TCP mode, which is set by adding mode tcp, you should also add option tcplog. With this option, the log format defaults to a structure that provides useful information like Layer 4 connection details, timers, byte count and so on.

Below is example of configured logging with some explanations:

Log-format "%ci:%cp [%t] %ft %b/%s %Tw/%Tc/%Tt %B %ts %ac/%fc/%bc/%sc/%rc %sq/%bq"

haproxy-logged-fields-explained
Example of Log-Format configuration as shown above outputted of haproxy config:

Log-format "%ci:%cp [%tr] %ft %b/%s %TR/%Tw/%Tc/%Tr/%Ta %ST %B %CC %CS %tsc %ac/%fc/%bc/%sc/%rc %sq/%bq %hr %hs %{+Q}r"

haproxy_http_log_format-explained1

To understand meaning of this abbreviations you'll have to closely read  haproxy-log-format.txt. More in depth info is to be found in HTTP Log format documentation


haproxy_logging-explained

Logging HTTP request headers

HTTP request header can be logged via:
 

 http-request capture

frontend website
    bind :80
    http-request capture req.hdr(Host) len 10
    http-request capture req.hdr(User-Agent) len 100
    default_backend webservers


The log will show headers between curly braces and separated by pipe symbols. Here you can see the Host and User-Agent headers for a request:

192.168.150.1:57190 [20/Dec/2018:22:20:00.899] website~ webservers/server1 0/0/1/0/1 200 462 – – —- 1/1/0/0/0 0/0 {mywebsite.com|Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Ubuntu Chromium/71.0.3578.80 } "GET / HTTP/1.1"

 

Haproxy Stats Monitoring Web interface


Haproxy is having a simplistic stats interface which if enabled produces some general useful information like in above screenshot, through which
you can get a very basic in browser statistics and track potential issues with the proxied traffic for all configured backends / frontends incoming outgoing
network packets configured nodes
 experienced downtimes etc.

haproxy-statistics-report-picture

The basic configuration to make the stats interface accessible would be like pointed in above config for example to enable network listener on address
 

https://192.168.0.5:8080/stats


with hproxyuser / password config would be:

# HAProxy Monitoring Config
#———————————————————————
listen stats 192.168.0.5:8080                #Haproxy Monitoring run on port 8080
    mode http
    option httplog
    option http-server-close
    stats enable
    stats show-legends
    stats refresh 5s
    stats uri /stats                            #URL for HAProxy monitoring
    stats realm Haproxy\ Statistics
    stats auth hproxyauser:Password___          #User and Password for login to the monitoring dashboard

 

 

Sessions states and disconnect errors on new application setup

Both TCP and HTTP logs include a termination state code that tells you the way in which the TCP or HTTP session ended. It’s a two-character code. The first character reports the first event that caused the session to terminate, while the second reports the TCP or HTTP session state when it was closed.

Here are some essential termination codes to track in for in the log:
 

Here are some termination code examples most commonly to see on TCP connection establishment errors:

Two-character code    Meaning
—    Normal termination on both sides.
cD    The client did not send nor acknowledge any data and eventually timeout client expired.
SC    The server explicitly refused the TCP connection.
PC    The proxy refused to establish a connection to the server because the process’ socket limit was reached while attempting to connect.


To get all non-properly exited codes the easiest way is to just grep for anything that is different from a termination code –, like that:

tail -f /var/log/haproxy.log | grep -v ' — '


This should output in real time every TCP connection that is exiting improperly.

There’s a wide variety of reasons a connection may have been closed. Detailed information about all possible termination codes can be found in the HAProxy documentation.
To get better understanding a very useful reading to haproxy Debug errors with  is in haproxy-logging.txt in that small file are collected all the cryptic error messages codes you might find in your logs when you're first time configuring the Haproxy frontend / backend and the backend application behind.

Another useful analyze tool which can be used to analyze Layer 7 HTTP traffic is halog for more on it just google around.