jrosa770/carp_haproxy

Redundant HaProxy with CARP Based Dynamic Fail Over

Redundant HaProxy with CARP Based Dynamic Fail Over

Audience

Network and/ or System Administrators

Pre-requisites: Basic understanding of IPv4 concepts and routing and understanding of TCP load balancing. Basic UNIX shell knowledge. Familiarity with HAProxy or similar load balancing configurations.

Motivation:

A solution that could provide:

1. Load balancing for MySQL (Percona in my case) and basic HTTP and HTTPS if needed
2. Redundancy and a method for dynamic fail over.
3. Configuration Management and Synchronization between primary and secondary load balancers

Sure I can have a vendor like F5 or A10 provide the functionality out of the box. But where is the challenge on that? Why not create it possible? What is my budget does not allow for the expense on a vendor based appliance? Open Source is there for a reason... Why not use it?

Solution

The solution I settled for was based on FreeBSD UNIX using CARP a method for failover an redundancy similar to Cisco HSRP or the open standard VRRP. For Load Balancing the best solution I've seen is HA Proxy, a fantastic TCP based load balancer.

Method

The following example is based on that solution. The basic setup requires two FreeBSD boxes or as VM(s). If VM(s) the recommendation is for two guests in two different hosts systems. An of course a network or set of available network and last bu not least the end systems that will ultimately handle the user's request for services.

Basic Steps

(See the full How To File in How To Documents )

1. First enable IP routing and CARP on the HAProxy Systems by configuring /etc/sysctl.conf on both systems

2. Configure the IP's and CARP Groups. If you're familiar with VRRP or HSRP this part will look very familiar as the basics are covered with a redundancy group with an ID as a number and a Virtual IP attached to that group. /etc/rc.conf (2 files included for Primary and Secondary)

   2(a). File rc.conf-primary to /etc/rc.conf on primary

   2(b). File rc.conf-secondary to /etc/rc.conf on secondary

3. Configure the HAProxy Daemon

   3(a). Configure the /usr/local/etc/haproxy.conf file on the Primary HAProxy Daemon with primary_haproxy.conf

   3(b). Configure the the /usr/local/etc/haproxy.conf file on the Secondary HAProxy Daemon with secondary_haproxy.conf

4. This setup requires a separate probe for Health checks at the destination or HaProxy Defined backend server(s). In this case MySQL. /var/lib/mysql-check/mysqlchk.mysq.bash

5. Configure the Service for Port 9200 TCP Health Check at the destination or HaProxy Defined backend server(s) -- Xinetd Service for TCP 9200 -- /etc/xinetd.d/mysqlcheck

The probe file can be adapted to monitor other services using either standard or non-standard TCP ports

Management Scripts

HaSync

HaSync is Bash script that builds the haproxy.conf files from sections. The intention is rather than editing the main haproxy.conf file itself, the administrator can add or edit then stage the different sections., depending on need or personal preference. This is specially useful on larger implementations, issue isolation and ease of use. The script will then gather the sections into a single haproxy.conf file to then sync the configuration from the primary HAProxy to the Secondary (Requires the etckeeper and diffcolor packages). A second hasync file named hasync_from_primary is installed on the standby for proper synchronization.

For our example the script assumes following directory structure:

#/usr/local/etc/haproxy/conf.d/

#├── global.cfg
#├── stats.cfg
#├── pcl.cfg (Percona/ MySQL)
#├── http.cfg 
#├── https.cfg
#├── others.cfg (Any Non Standard TCP Port)

Every site has it's own file, so you can disable site by changing it's file extension, or appending .disabled. You can add ad many files as needed to create as many sections you require in your haproxy.conf file. Changes or additions can be done via console, via scp from an external source, scripting combinations or orchestration like Ansible.

HaConf

HaConf makes a Backup of the current haproxy.conf file. Then the current haproxy file is opened in ee (FreeBSD Easy Editor) for editing.

The editor can be changed from ee to vi or any other editor. (Line 13 in haconf: from ee $Original to vi $Original) The script is intended as a standalone editing routing but does not provide for synchronization with the standby. If synchronization is needed the better alternative is hasync. Haconf is intended mainly for initial configurations, test or standalone HAProxy setups.