We have some CCR1016s running PPPoE servers, handling up to 800 sessions in one case. If the CCR reboots or the uplink to the RADIUS server etc. is lost, when things go back to normal, all clients want to connect, all at once.
The RADIUS server is thus hit with hundreds of access & accounting requests. We have been able to limit accounting request volumes by adding a random Acct-Interim-Interval to each client profile, see before & after (RADIUS hit with simultaneous accounting updates vs. randomized times):
I have read some posts that suggest using Bridge filters to limit PPPoE requests, but is this not going to have a CPU hit on the CCR? We are already seeing timeouts on occasion, which we think could be the CCR reaching its limits.
Any suggestions welcome!