Autobandwidth is supported for LSPs that have secondary or secondary standby paths. A secondary path is only initialized at its configured bandwidth when it is established, and the bandwidth is adjusted only when the secondary path becomes active.
This description makes use of the following terminology:
current_BW: the last known reserved bandwidth for the LSP; may be the value of a different path from the currently active path
operational BW: the last known reserved BW for a specified path, as recorded in the MIB
configured BW: the bandwidth explicitly configured for the LSP path by the user in CLI
active path: the path (primary or secondary) the LSP currently uses to forward traffic
signaled BW: the new BW value signaled during an MBB
A secondary or standby secondary path is initially signaled with its configured bandwidth. Setup for the secondary path is triggered only when the active path goes down or becomes degraded (for example, because of FRR or preemption). An auto-BW triggered bandwidth adjustment (auto bandwidth MBB) only takes place on the active path. For example, if an auto-BW adjustment occurs on the primary path, which is currently active, no adjustment is made at that time to the secondary path because that path is not active.
When the active path changes, the current_bw is updated to the operational bandwidth of the newly active path. While the auto-BW MBB on the active path is in progress, a statistics sample could be triggered, and this would be collected in the background. Auto-bandwidth computations use the current_bw of the newly active path. In case the statistics sample collection results in a bandwidth adjustment, the in-progress auto-BW MBB is restarted. If after five attempts, the auto-BW MBB fails, the current_bw and secondary operational BW remain unchanged.
For a secondary or standby secondary path, if the active path for an LSP changes (without the LSP going down), an auto-BW MBB is triggered for the new active path. The bandwidth used to signal the MBB is the operational bandwidth of the previous active path. If the MBB fails, it retries with a maximum of five attempts. The reserved bandwidth of the newly active path is therefore its configured bandwidth until the MBB succeeds.
For a secondary path where the active path goes down, the LSP goes down temporarily until the secondary path is setup. If the LSP goes down, all statistics and counters are cleared, so the previous path operational bandwidth is lost. That is, the operational BW of a path is not persistent across LSP down events. In this case, there is no immediate bandwidth adjustment on the secondary path.
The following algorithm is used to determine the signaled bandwidth on a newly active path:
For a path that is operationally down, signaled_bw = config_bw.
For the active path, if an auto-BW MBB adjustment is in progress, signaled_bw = previous path operational BW for the first five attempts. For the remaining attempts, the signaled BW = operational BW.
For an MBB on the active path (other than an auto-BW MBB), MBB signaled BW = operational BW.
For an MBB on the inactive path, MBB signaled BW = configured BW.
If the primary path is not the currently active path and it has not gone down, then any MBB uses the configured BW for the primary path. However, if the configured BW is changed for a path that is currently not active, then a config change MBB is not triggered.
If the standby is SRLG enabled, and the active path is the standby, and the primary comes up, this immediately triggers a delayed retry MBB on the standby. If the delayed retry MBB fails, immediate reversion to the primary occurs regardless of the retry timer.
When the system reverts from a secondary standby or secondary path to the primary path, a Delayed Retry MBB is attempted to bring bandwidth of the standby path back to its configured bandwidth. Delayed Retry MBB is attempted once, and if it fails, the standby is torn down. A Delayed Retry MBB has highest priority among all MBBs, so it takes precedence over any other MBB in progress on the standby path (for example, Config change or Preemption).
The system carries over the last signaled BW of the LSP over multiple failovers. For example, if an LSP is configured with auto-BW for some time, and adjusts its currently reserved bandwidth for the primary, and Monitor mode is then enabled, BW adjustment on the primary ceases, but the BW remains reserved at the last adjusted value. Next, the LSP fails over to a secondary or secondary standby. The secondary inherits the last reserved BW of the primary, but then disables further adjustment as long as monitoring mode is enabled.
The system’s ability to carry-over the last signaled BW across failovers has the following limitations:
Case 1 — If the LSP fails over from path1 to path2 and the AutoBW MBB on path2 is successful, the last signaled BW is carried over when the LSP reverts back to path1 or fails over to a new path3. This may trigger an AutoBW MBB on the new active path to adjust its bandwidth to last signaled BW.
Case 2 — If the LSP fails over from path1 to path2 and the AutoBW MBB on path2 is still in progress and the LSP reverts back to path1 or fails over to a new path3, the last signaled BW is carried over to the new active path (path1 or path3) and this may result in an AutoBW MBB on that path.
Case 3 — If the LSP fails over from path1 to path2 and the AutoBW MBB on path2 fails (after 5 retry attempts), the last signaled BW from when path1 was active is lost. Therefore, when the LSP reverts back to path1 or fails over to a new path3, the original signaled BW from path1 is not carried over. However the signaled bandwidth of path2 is carried over to the new active path (path1 or path3) and may trigger an AutoBW on that path.