Geolocation-based load balancing

Install and enable the GSLB service

You can store GeoIP database files, typically in a specific format like MaxMind's GeoIP2 or GeoLite2. GeoIP data is information about the geographical location of IP addresses. This data is used in GSLB to determine the optimal routing of client requests based on their geographic location. These database files contain mappings between IP addresses and their corresponding geographical information, such as country, region, city, and latitude/longitude coordinates.

When a client makes a request to the GSLB system, the system can analyze the client's IP address and consult the GeoIP database to determine the client's location. Based on this information, the GSLB system can make intelligent routing decisions to direct the client's request to the most appropriate server or data center that can serve the request efficiently and optimize network performance.

Geobalancing using MaxMind

In order to use Geobalancing in your environment, you will need to download a GeoIP database via MaxMind.

1. Create your account through the MaxMind website and download the GeoIP databases.

2. Copy your geolocation database files to a chosen data directory through a utility such as scp or Winscp. For example, store City data at /data/GeoLite2-City.mmdb:

   $ scp GeoLite2-City.mmdb admin@172.16.24.238:/data/geoip/

3. In your zone file, /etc/hapee-extras/hapee-gslb.conf, include an additional record of type map to use your MaxMind GeoIP2 City database to change the DNS response to contain a datacenter's IP based on the client's location. The map record type refers to a geoip-map section with the same name.

   In the example below, we map the domain www.example.com to the mymap geoip-map section, which uses the geolocation database to return different IP addresses in response to DNS queries depending on whether the client is in France or the United States.

   zone example.com
     ttl 84600
     record @ ttl 900  SOA   ns1 hostmaster 1 7200 30M 3D 900
     record @          NS    ns1.example.com.
     record ns1        A     203.0.113.1   # nameserver: HAProxy IP address
     record alias      CNAME www
     record www        map   mymap
   
   geoip-map mymap
     location-base /data/geoip/GeoLite2-City.mmdb
     location EU/FR/Paris DC1 DC2
     location NA/US/Chicago DC2 DC1
     network 198.51.100.0/24 DC1 DC2
     network 203.0.113.0/24 DC2 DC1

   The geoip-map section directives are as follows:

   geoip-map section directive	Description	Example
   location-base	Absolute path to the geolocation database. You can supply several geolocation database names separated by spaces.	location-base /data/geoip/GeoLite2-City.mmdb
   location	The first parameter is a hierarchical path to a geographic region in the order of the continent code, a country ISO code, then more specific regions like state and city name. Refer to the MaxMind reference guide and ISO-3166 for these codes. Note that GSLB will search deeper into the hierarchy if a match is not found at the current layer. For example, you could specify country and city name, but omit the state name between them. The second parameter is a space-separated list of answer-list section names (e.g. DC2). GSLB directs client requests sent from this location to the first healthy datacenter in the list.	location NA/US/NY DC2
   network	As an alternative to using location, which uses geolocation data to choose the datacenter, you can also specify a client IP range. Set a subnet value in CIDR notation followed by an ordered list of datacenters (separated by spaces). The second parameter is a space-separated list of answer-list section names (e.g. DC2). GSLB directs client requests sent from this subnet to the first healthy datacenter in the list.	network 198.51.100.0/24 DC1

4. Create new answer-list sections.

   The answer-list for DC1 contains a list of IP addresses for the datacenter in Europe, while the DC2 answer-list contains a list of IP addresses for the datacenter in North America.

   zone example.com
     ttl 84600
     record @ ttl 900  SOA   ns1 hostmaster 1 7200 30M 3D 900
     record @          NS    ns1.example.com.
     record ns1        A     203.0.113.1   # nameserver: HAProxy IP address
     record alias      CNAME www
     record www        map   mymap
   
   geoip-map mymap
     location-base /var/run/hapee-extras/gslb/geoip/GeoIP2-City.mmdb
     location EU/FR/Paris DC1 DC2
     location NA/US/Chicago DC2 DC1
     network 198.51.100.0/24 DC1 DC2
     network 203.0.113.0/24 DC2 DC1
   
   answer-list DC1
     up_threshold 0.5
     method single-rr
     option tcpchk fall 10 rise 10
     tcp-check connect port 80
     answer-record srv1 198.51.100.1 weight 20
     answer-record srv2 198.51.100.2 weight 20
     answer-record srv3 198.51.100.3 weight 10
     answer-record srv4 2001:db8::4001 weight 20
     answer-record srv5 2001:db8::4002 weight 20
     answer-record srv6 2001:db8::4003 weight 10
   
   answer-list DC2
     up_threshold 0.5
     method single-rr
     option httpchk
     http-check connect
     http-check send uri /health.html hdr host www.example.com
     http-check expect status 200,301,302
     answer-record srv1 203.0.113.10  weight 20
     answer-record srv2 203.0.113.11  weight 20
     answer-record srv3 203.0.113.12  weight 10

   GSLB will send DNS responses based on the location of the client. It will only send either IPv4 or IPv6 addresses, depending on the type of IP addresses the client requests.

   Weights determine how often a particular IP address will be returned, with higher weights being chosen more often. The weight values apply only to the IPv4 or IPv6 pool of IP addresses. In the example, the IPV4 pool of servers and the IPv6 pool of servers have their own total weight sums.

   By specifying method single-rr, HAProxy Enterprise alternates which IP for a datacenter it sends to clients in order to distribute traffic across all servers. In that case, the odds of a server's IP being returned is server weight / sum of all server weights.

   You can set method multi-rr to return multiple IP addresses to the client. In that case, the odds of a server's IP being returned is server weight / max weight value.

   The up_threshold directive determines the percentage of servers that must be up. Otherwise, traffic is routed to a different datacenter altogether.

5. Save your configuration, close the configuration editor, then apply your changes. Enable and start the GSLB service.

   $ sudo systemctl restart hapee-extras-gslb

Testing

When testing GSLB Geo Balancing, Dig will provide a response from a healthy server that best matches the set geographic preferences. If you have configured your GSLB system to prioritize servers located in the same country as the client, Dig will return a response from a healthy server in that country. For example, if your domain is example.com you can use dig A @127.0.0.1 -p 153 example.com to test.

; <<>> DiG 9.16.1-Ubuntu <<>> example.com @192.168.1.100
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 1809
;; flags: qr rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 1

;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags:; udp: 512
;; QUESTION SECTION:
;example.com.              IN      A

;; ANSWER SECTION:
example.com.        3600    IN      A       93.184.216.34

;; Query time: 50 msec
;; SERVER: 192.168.1.100#53(192.168.1.100)
;; WHEN: Mon Jul 10 17:02:00 UTC 2023
;; MSG SIZE  rcvd: 57