Reference
add backend
Available since
- HAProxy 3.4
Add a dynamic backend.
You can add, publish, unpublish, and delete dynamic backends at runtime without performing a reload.
To ensure that named defaults sections are available to dynamic backends, they’re kept in memory. If you don’t use dynamic backends, you can save memory by disabling this feature. To purge defaults sections from memory, set the global directive tune.defaults.purge.
This command requires a CLI level of admin. Note that this operation is supported only on a CLI connection running in experimental mode (experimental-mode on). This operation is still in development and may change in the future.
Description Jump to heading
A backend inherits its settings from the defaults section specified with the from argument (for example, backend mybackend from mydefaults). If the defaults section doesn’t specify a mode, set one using the mode argument. The add backend command only adds the backend to the load balancer’s runtime memory and not to the load balancer configuration saved on disk.
Initially, the backend is unavailable to receive traffic. To make it available for traffic, use publish backend. Add servers to the backend using add server. If a server is configured for health checks, the health checks are active as soon as the server is added to the backend, even if the backend hasn’t yet been published.
If you target a backend with the default_backend or use_backend directive in a frontend and then you disable or unpublish that backend, it’ll be skipped as a default backend. However, you can force the load balancer to use the backend if you set the force-be-switch directive.
Tip
You can force the load balancer to use the backend even if it’s disabled or unpublished by using the force-be-switch directive. This directive can be used by admins to test traffic to services prior to exposing them to the outside world.
Examples Jump to heading
In this example, we add a dynamic backend named mybackend1 using defaults from section mydefaults.
Here’s the defaults section mydefaults:
haproxydefaults mydefaultstimeout http-request 5stimeout connect 5s timeout client 30s timeout server 10s
haproxydefaults mydefaultstimeout http-request 5stimeout connect 5s timeout client 30s timeout server 10s
The defaults section doesn’t specify the mode, so we include the mode argument in the add backend command. Be sure to turn on experimental mode.
nixecho "experimental-mode on; add backend mybackend1 mode http from mydefaults" | \sudo socat stdio tcp4-connect:127.0.0.1:9999
nixecho "experimental-mode on; add backend mybackend1 mode http from mydefaults" | \sudo socat stdio tcp4-connect:127.0.0.1:9999
outputtextNew backend registered.
outputtextNew backend registered.
Here’s the resulting backend:
haproxybackend mybackend1 from mydefaultsmode http
haproxybackend mybackend1 from mydefaultsmode http
Add a backend, server, and map file entry Jump to heading
In the following example, we add a backend, a server, and a map file entry that directs traffic to the new backend. Then we publish the backend so it becomes available for use in the load balancer. For a tutorial on map files, see Map files.
Before we make any changes, our example load balancer configuration starts off like this:
- In the global section, we enable the HAProxy Runtime API at port 9999.
- There’s a defaults section named
mydefaults. - There’s a frontend named
myfrontend1. - The
use_backenddirective reads a map file to determine the correct backend to use based on the requested URL path. The map file is virtual, meaning it exists in memory, and starts off empty. - If there isn’t a match in the map file for the requested URL path, the
default_backenddirective targets traffic to the backend namedwebservers.
Here’s the configuration file:
haproxyglobalstats socket ipv4@127.0.0.1:9999 level admindefaults mydefaultslog globalmode httpoption httplogoption dontlognulltimeout connect 10mtimeout client 10mtimeout server 10mfrontend myfrontend1bind :80use_backend %[path,map_beg(virt@paths.map)]default_backend webserversbackend webserversserver web1 127.0.0.1:8080
haproxyglobalstats socket ipv4@127.0.0.1:9999 level admindefaults mydefaultslog globalmode httpoption httplogoption dontlognulltimeout connect 10mtimeout client 10mtimeout server 10mfrontend myfrontend1bind :80use_backend %[path,map_beg(virt@paths.map)]default_backend webserversbackend webserversserver web1 127.0.0.1:8080
To add our dynamic backend and make the other changes, follow these steps:
-
Add a dynamic backend using the
add backendcommand. The backend is namedmybackend1and uses defaults from sectionmydefaults.nixecho "experimental-mode on; add backend mybackend1 from mydefaults" | \sudo socat stdio tcp4-connect:127.0.0.1:9999nixecho "experimental-mode on; add backend mybackend1 from mydefaults" | \sudo socat stdio tcp4-connect:127.0.0.1:9999 -
Add a server using the
add servercommand.nixecho "add server mybackend1/web1 172.16.0.12:8080 check port 8080" | \sudo socat stdio tcp4-connect:127.0.0.1:9999nixecho "add server mybackend1/web1 172.16.0.12:8080 check port 8080" | \sudo socat stdio tcp4-connect:127.0.0.1:9999 -
Enable the server using the
enable servercommand.nixecho "enable server mybackend1/web1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999nixecho "enable server mybackend1/web1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999 -
Enable health checks for the server using the
enable healthcommand.nixecho "enable health mybackend1/web1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999nixecho "enable health mybackend1/web1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999 -
Make the backend available for traffic using the
publish backendcommand.nixecho "publish backend mybackend1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999nixecho "publish backend mybackend1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999 -
Add an entry to the map file.
nixecho "add map virt@paths.map /test mybackend1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999nixecho "add map virt@paths.map /test mybackend1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999
The backend is available to accept traffic, and the map file has been updated with an entry that routes requests for the URL path /test to the new backend. Note that both the dynamic backend and the virtual map file reside only in memory in the running load balancer process. They do not exist in the configuration file on disk.
Delete backend, server, and map file entry Jump to heading
In the following example, we delete the server, backend, and map entry. Use these commands:
-
Put the server into maintenance mode. This command marks the server as “DOWN” for maintenance. The load balancer stops sending checks to this server while it’s in this state.
nixecho "set server mybackend1/web1 state maint" | \sudo socat stdio tcp4-connect:127.0.0.1:9999nixecho "set server mybackend1/web1 state maint" | \sudo socat stdio tcp4-connect:127.0.0.1:9999 -
Wait for the server to be removable and then remove it. A server is removable when it no longer has any active or idle connections.
nixecho "wait 2s srv-removable mybackend1/web1; del server mybackend1/web1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999nixecho "wait 2s srv-removable mybackend1/web1; del server mybackend1/web1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999 -
Unpublish the backend.
nixecho "unpublish backend mybackend1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999nixecho "unpublish backend mybackend1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999 -
Wait for the backend to be removable and then delete it.
nixecho "experimental-mode on; wait 2s be-removable mybackend1; del backend mybackend1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999nixecho "experimental-mode on; wait 2s be-removable mybackend1; del backend mybackend1" | \sudo socat stdio tcp4-connect:127.0.0.1:9999 -
Delete the map file entry.
nixecho "del map virt@paths.map /test" | \sudo socat stdio tcp4-connect:127.0.0.1:9999nixecho "del map virt@paths.map /test" | \sudo socat stdio tcp4-connect:127.0.0.1:9999
See also Jump to heading
- del backend
- del map
- del server
- experimental mode
- publish backend
- set server
- unpublish backend
- wait
- To target a specific backend, see Use conditionals to forward traffic to different backends.
- To direct traffic to a backend even when disabled or unpublished, use the force-be-switch directive.
- If you don’t use dynamic backends, consider setting tune.defaults.purge.