S-Series - High CPU Utilization - Many PIM neighbors out one port - tRtrPtcls task
High CPU Utilization on Fabric blade
tRtrPtcls task consuming more CPU than usual
Dropping PIM neighbors
Losing multicast streams
Many PIM neighbors on the other side of a single link
Example: Router with a single connection to a Metro-E that connects to many other routers.
PIM is processing too many protocol messages on one link. Internal rate limiters cause us to drop some messages and the PIM table ends up thrashing, driving up the CPU.
Upgrade to 8.62.03
The fix in firmware was to give PIM its own queue and increase the ratelimit for sending these packets to the host.
Note: This fix was designed to keep the neighbor state alive to a large number of PIM neighbors out one link. If CPU is being driven up for another (possibly legitimate) reason, this firmware may have no impact on a high CPU utilization scenario.