Supported Platforms
Configuring Back-up Options for RADIUS Accounting
You can configure RADIUS accounting backup to preserve accounting data when the accounting server is unavailable because of a server or network outage. When backup is configured, RADIUS accounting stop messages are withheld and queued to be sent when connectivity is restored. You can specify the maximum number of stop messages that can be queued. When this maximum is reached, subsequent new subscriber logins fail because there is no remaining capacity to preserve accounting data for new sessions.
You can also configure how long the queued messages can be held. When this period expires, all pending accounting stops are flushed from the queue, even if the accounting server has come back online.
![]() | Note: Configuring accounting backup disables the revert timer. An error message is generated if you attempt to configure the revert-interval statement at the [edit access profile profile-name options] or [edit access radius-options] hierarchy levels. |
![]() | Caution: Before you configure RADIUS accounting backup, ensure that RADIUS accounting and RADIUS authentication are configured on different servers. Subscriber authentication fails when the same server is configured for both authentication and accounting. |
- Enable accounting backup to use the default values.[edit access ]user@host# set accounting-backup-options
- (Optional) Configure the number of accounting stops that
the router can preserve while the accounting server is offline.[edit access accounting-backup-options]user@host# set max-pending-accounting-stops number
- (Optional) Configure how long the router holds pending
accounting stops before flushing them.[edit access accounting-backup-options]user@host# set max-withhold-time hold-time
For example, the following statements configure the backup options for all subscriber accounting; these statements specify that the router holds no more than 32,000 pending accounting stops—at which point all subsequent subscriber logins fail—and holds them no longer than 6 hours—at which point all pending messages are flushed and subscriber logins resume if they were failing: