Description

This article informs about the trace-options configured on the MX204 router, could increase high CPU utilisation on the router.

Symptoms

When we see log messages such as "JTASK_SCHED_SLIP" on the router, it denotes the PFE/Kernal is busy and it wasn’t execute the event within the time.

 

Aug 28 08:31:32 mx204-re0 rpd[12436]: JTASK_SCHED_SLIP: 4 sec scheduler slip, user: 0 sec 129397 usec, system: 1 sec, 285932 usec

 

We can also cross check this higher utilised process on the device, by executing the command "show system processes extensive"

 

root@router> show system processes extensive no-forwarding

last pid: 25540; load averages: 1.63, 1.51, 1.76 up 4+23:11:36  08:41:06

476 threads: 6 running, 421 sleeping, 49 waiting

CPU: 2.3% user, 0.0% nice, 24.1% system, 0.2% interrupt, 73.5% idle

Mem: 919M Active, 4097M Inact, 1024M Wired, 304M Buf, 10G Free

Swap: 3072M Total, 3072M Free

 

 PID USERNAME  PRI NICE SIZE  RES STATE  C TIME  WCPU COMMAND

  11 root    155 ki31  0B  80K CPU0  0 86.9H 50.76% idle{idle: cpu0}

  11 root    155 ki31  0B  80K CPU2  2 89.3H 55.49% idle{idle: cpu2}

  11 root    155 ki31  0B  80K RUN   1 86.8H 60.78% idle{idle: cpu1}

  11 root    155 ki31  0B  80K CPU4  4 86.1H 53.85% idle{idle: cpu4}

  11 root    155 ki31  0B  80K CPU3  3 87.0H 60.29% idle{idle: cpu3}

12436 root    23  0 2722M 1842M kqread 2 19.8H 36.49% rpd{bgpio-0-th}

12436 root    29  0 2722M 1842M biowr  3 22.4H 30.58% rpd{TraceThread}

12436 root    49  0 2722M 1842M kqread 0 22.0H 18.98% rpd{rpd}

  34 root    -16  -  0B  48K psleep 3 809:34 17.67% pagedaemon{dom0}

 

We could see the processes 'rpd{bgpio-0-th}, rpd{TraceThread}, and rpd, pagedaemon' are utilized higher.

RPD trace-thread will be utilising higher, if the trace-options are configured:

 

system {

processes {

    dhcp-service {

      traceoptions {

        file dhcplogfile size 10m;

        level all;

        flag packet;

      }

    }

  }

  

 

routing-instances {

  instance-name {

    protocols {

      bgp {

        traceoptions {

          file instance-bgp;

          flag all;

        }

Solution

  • Trace-options would consume more CPU at the background and it is advised that trace-options are to be disabled, once the necessary logs are collected.
  • By deactivating the below trace-options configured on the router, the CPU process utilisation would be reduced. 

#deactivate system dhcp-service traceoptions

#deactivate routing-instances instance-name protocols bgp traceoptions

#commit

Modification History

2024-09-06 : Article Created