Chapter 5. HAProxy Configuration
This chapter explains the configuration of a basic setup that highlights the common configuration options an administrator could encounter when deploying HAProxy services for high availability environments.
HAProxy has its own set of scheduling algorithms for load balancing. These algorithms are described in Section 5.1, “HAProxy Scheduling Algorithms”.
HAProxy is configured by editing the
Load Balancer configuration using HAProxy consists of five sections for configuration:
- The proxies section, which consists of 4 subsections:
5.1. HAProxy Scheduling Algorithms
The HAProxy scheduling algorithms for load balancing can be edited in the
balance parameter in the
backend section of the
/etc/haproxy/haproxy.cfg configuration file. Note that HAProxy supports configuration with multiple back ends, and each back end can be configured with a scheduling algorithm.
- Round-Robin (
- Distributes each request sequentially around the pool of real servers. Using this algorithm, all the real servers are treated as equals without regard to capacity or load. This scheduling model resembles round-robin DNS but is more granular due to the fact that it is network-connection based and not host-based. Load Balancer round-robin scheduling also does not suffer the imbalances caused by cached DNS queries. However, in HAProxy, since configuration of server weights can be done on the fly using this scheduler, the number of active servers are limited to 4095 per back end.
- Static Round-Robin (
- Distributes each request sequentially around a pool of real servers as does Round-Robin, but does not allow configuration of server weight dynamically. However, because of the static nature of server weight, there is no limitation on the number of active servers in the back end.
- Least-Connection (
- Distributes more requests to real servers with fewer active connections. Administrators with a dynamic environment with varying session or connection lengths may find this scheduler a better fit for their environments. It is also ideal for an environment where a group of servers have different capacities, as administrators can adjust weight on the fly using this scheduler.
- Source (
- Distributes requests to servers by hashing requesting source IP address and dividing by the weight of all the running servers to determine which server will get the request. In a scenario where all servers are running, the source IP request will be consistently served by the same real server. If there is a change in the number or weight of the running servers, the session may be moved to another server because the hash/weight result has changed.
- URI (
- Distributes requests to servers by hashing the entire URI (or a configurable portion of a URI) and divides by the weight of all the running servers to determine which server will the request. In a scenario where all active servers are running, the destination IP request will be consistently served by the same real server. This scheduler can be further configured by the length of characters at the start of a directory part of a URI to compute the hash result and the depth of directories in a URI (designated by forward slashes in the URI) to compute the hash result.
- URL Parameter (
- Distributes requests to servers by looking up a particular parameter string in a source URL request and performing a hash calculation divided by the weight of all running servers. If the parameter is missing from the URL, the scheduler defaults to Round-robin scheduling. Modifiers may be used based on POST parameters as well as wait limits based on the number of maximum octets an administrator assigns to the weight for a certain parameter before computing the hash result.
- Header Name (
- Distributes requests to servers by checking a particular header name in each source HTTP request and performing a hash calculation divided by the weight of all running servers. If the header is absent, the scheduler defaults to Round-robin scheduling.
- RDP Cookie (
- Distributes requests to servers by looking up the RDP cookie for every TCP request and performing a hash calculation divided by the weight of all running servers. If the header is absent, the scheduler defaults to Round-robin scheduling. This method is ideal for persistence as it maintains session integrity.