This article describes the issue of subscriber login failure when hierarchical CoS has applied on MX device.
When customer enabled hierarchical CoS on subscriber interface, the subscriber may login failure due to Cos default resource exhausted.
Enable hierarchical CoS refer below example:
set dynamic-profiles pppoe-profile class-of-service traffic-control-profiles TC_PROF_TRAFFICSHAPER scheduler-map default-sch-01
set dynamic-profiles pppoe-profile class-of-service traffic-control-profiles TC_PROF_TRAFFICSHAPER shaping-rate "$junos-cos-shaping-rate"
set dynamic-profiles pppoe-profile class-of-service traffic-control-profiles TC_PROF_TRAFFICSHAPER shaping-rate burst-size "$junos-cos-shaping-rate-burst"
set dynamic-profiles pppoe-profile class-of-service interfaces pp0 unit "$junos-interface-unit" output-traffic-control-profile TC_PROF_TRAFFICSHAPER
Hierarchical CoS for Subscriber Interfaces refer below link:
https://www.juniper.net/documentation/us/en/software/junos/cos-hierarchical/topics/concept/hierarchical-scheduling-mic-mpc-understanding.html
Using below cli to check Cos resource on each PFE:
user@mx# run show system resource-monitor summary
(Full output truncated)
CoS Queue Utilization
PFE # Scheduler Block # Used %
0 0 160 0
1 0 16384 100 * <<<<<PFE CoS Queue Utilization has reached to 100%
2 0 96 0
3 0 96 0
Set fpc max-queues to maximum value, the pfe cos queue resource can expand to platform maximum value automatically. Once this maximum value has been used by 100%, that means it reaches hardware limitation.
Refer below MX304 example:
# set chassis fpc 0 max-queues 2M
user@mx# run show system resource-monitor summary(Full output truncated) CoS Queue Utilization PFE # Scheduler Block # Used % 0 0 160 0 1 0 32176 49 <<<<<PFE CoS Queue Utilization expands from 16k to 32k 2 0 96 0 3 0 96 0
ATTENTION: modify this value can cause FPC rebooting.