Customer reported a high CPU utilization issue, with the mcsnoopd daemon spiking to 100%
mcsnoopd
We observed high utilization of the mcsnoopd daemon from the below output, which also appears to have caused an increase in rpd utilization
rpd
labroot> show system process extensive
PID USERNAME PRI NICE SIZE RES STATE C TIME WCPU COMMAND
71621 root 102 0 798M 43M CPU1 1 316.4H 100.00% mcsnoopd{mcsnoopd}
11747 root 91 0 1411M 522M CPU0 0 218.6H 63.38% rpd{rpd}
All Layer 2 services were migrated from the old router to the current router, and the issue encountered following the migration. The deamon mcsnoopd was 0.00% before migrating the L2services to the router , after migration it went upto 100% and we could also observe even rpd spike to 63.38%.
The device is hitting PR1710565.
Before Migration :
71621 root 20 0 794M 42M kqread 2 88:50 0.00% mcsnoopd{TraceThread}
11747 root 20 0 1407M 520M kqread 0 25.6H 0.00% rpd{rpd}
After Migration :
According to the PR following are the triggers that can cause the device to hit the PR.
On all Junos and Junos Evolved platforms, whenever a commit is done, that involves mcsnoopd daemon config parsing such as (VLAN creation/deletion, interface add/delete to VLAN, interface enable/disable, IGMP (Internet Group Management Protocol) snooping/MLD (Multicast Listener Discovery) snooping related config commands) mcsnoopd will consume CPU and it may affect other daemons like rpd. In the subjected case, it was observed that the addition of VLAN and IGMP-related configurations were the triggers for the issue.
This issue has been resolved in the following Junos versions; however, the customer was running a Junos version in which the issue was not fixed
Resolved-In
evo:22.2R3-S5-EVO evo:22.3R3-S4-EVO evo:23.2R2-S2-EVO evo:23.4R2-EVO junos:21.2R3-S9 junos:22.2R3-S5 junos:22.3R3-S4 junos:22.4R3-S2 junos:23.2R2-S2 junos:23.4R2 junos:24.1R1
Workaround
Delete the igmp-snooping configuration