If I define more than one Backup Service, is traffic distributed across all of them?

Short answer

No. When all the main Backend Services are unavailable, only the first available Backup Service in the order receives the traffic; the other backups stay on standby. If the first backup also becomes unavailable (if it falls into the Error status in the health check or is put into Maintenance Mode), the next backup takes over. The order is determined by the order of the Default Backend Services list on the vService and, in a Backend Group, by the row order in the Backends table. This behavior applies on HTTP and L7 TCP vServices.

How Does It Work?

• TR7 does not distribute load among the backups; when all the main servers are unavailable, it uses only the first available backup.
• The "first" backup is determined according to the order written in the configuration:
  • the order of the Backend Services and groups in the list of the vService,
  • within a group, the row order in the Backends table (the group expands into its rows at the place where it is in the list).
• Each Conditional Backend Services set is evaluated separately; the backups of the set are selected according to the order in its own list.
• When a main Backend Service becomes OK again, new connections return to it; the connections that remain open on the backup are not cut, and the backups go back to standby.

Changing the Order

Where the backup is	How the order is changed
Default Backend Services list of the vService	The vService is opened with Edit, the selected items in the list are reordered by drag and drop, and Save is clicked
Backend Group (Backup Service on)	The group is opened with Edit, the rows in the Backends table are dragged by their handle, and it is saved with the Edit button at the bottom of the form

In TR7

• A backup is marked with the Backup Service switch in the Backend Service or in the Backend Group; its default value is off. In the list, an amber backup server icon appears in the name column.
• Which backup receives traffic is monitored from the current session counts on the Live Stats screen of the vService.
• While Clustering is on and Manual Sync Mode is off (default), the order and the backup setting are transferred to the peer device automatically.
• If Backup Service is on in a Backend Group, all the servers in the group are backups and follow the same rule: the first available row receives the traffic.
• The backup behavior is for HTTP and L7 TCP vServices; it is not applied on a Network vService.

Common Misconceptions

• "Two backups share the load when the main servers go down." Only the first available backup receives traffic.
• "A backup group is balanced like a server pool." In a backup group too, the traffic goes to the first available row.
• "The order is determined by the names of the Backend Services." The order is the placement order in the lists.

How it is done in the documentation

• Backend Services · Backup Service
• Backend Groups · Backup Service
• HTTP (L7) vService · Default Backend Services

Related questions

• What is a Backup Service; when does it take over, and when does it step back?
• For redundancy across data centers, should I use Load Balancer or GTM?
• When should I use a Backend Group, and when individual Backend Services?