[j-nsp] multicore JunOS CPU utilization collection method .. ?
Michael Hare
michael.hare at wisc.edu
Wed Dec 27 11:14:27 EST 2017
A bit of cli sleuthing and googling has left me blank, so before I spend time opening a jtac case...
In short, for multicore JunOS (MX 16.1, for example), does anyone have a method for retrieving per CPU core usage statistics? SNMP would be preferred, XML tolerable.
Historically I used 1.3.6.1.4.1.2636.3.1.13.1.8 [jnxOperatingCPU] to retrieve Routing Engine CPU usage, but in 16.X it looks like jnxOperatingCPU still only returns a single CPU value for the RE. To be clear, I am specifically interested in the following information(user/nice/system/interrupt summarized would be fine), which I can only get as far as I know interactively with the 'top' command from the shell
% top
last pid: 38153; load averages: 0.08, 0.28, 0.36 up 0+13:45:04 09:38:01
90 processes: 2 running, 87 sleeping, 1 zombie
CPU 0: 2.0% user, 0.4% nice, 4.7% system, 0.4% interrupt, 92.5% idle
CPU 1: 1.2% user, 0.4% nice, 5.9% system, 1.2% interrupt, 91.3% idle
CPU 2: 2.0% user, 0.0% nice, 3.5% system, 0.4% interrupt, 94.1% idle
CPU 3: 0.4% user, 0.0% nice, 3.1% system, 0.4% interrupt, 96.1% idle
Specifically, "show chassis routing-engine" still only displays a system total, not a per-core breakdown.
> show chassis routing-engine
Routing Engine status:
Slot 0:
...
5 sec CPU utilization:
User 0 percent
Background 0 percent
Kernel 0 percent
Interrupt 1 percent
Idle 99 percent
I do already grab load average, personal experience with runaway processes on multicore Cisco DSBU (Catalyst 38xx, for example) makes me want the per core CPU details.
-Michael
More information about the juniper-nsp
mailing list