Description

Customers observe 100% CPU utilization on EX4300 switch VCs. Process "eventd" is the major contributor which utilizes 50% or even 60% of total CPU usage.

This issue may be noticed after Junos upgrade activity to version 21.4R3-S5. CPU usage may spike to 100% for a short moment and drop to normal in the first following days after the upgrade. With time past, the CPU usage may keep running at 100% and never come down.

Symptoms

Example output:

root@EX4300-VC> show system processes extensive no-forwarding

last pid: 70685;  load averages:  2.31,  2.28,  2.17  up 41+18:50:08    18:15:32
184 processes: 6 running, 155 sleeping, 23 waiting

Mem: 1467M Active, 132M Inact, 151M Wired, 93M Cache, 112M Buf, 16M Free
Swap:


  PID USERNAME     PRI NICE   SIZE    RES STATE    TIME   WCPU COMMAND
 1326 root          76    0   294M   277M RUN    369.8H 31.74% eventd   <<<<<<<<<<<<<
 1800 root         -52  -52   642M   245M umtx   340.1H 17.63% pfex_junos
70620 root          64    0 97980K 55656K select   0:01 11.84% cli
 1800 root          76    0   642M   245M RUN    340.1H  8.01% pfex_junos
 1816 root          12    0     0K    16K peer_p 168.4H  1.95% ppt_11_80000010
70621 root          72    0 89908K  8476K select   0:00  1.75% mgd
 1794 root          52    0 74952K 37944K select  25.2H  1.71% chassisd
 1827 root          49    0 50724K 29892K select 282:52  0.83% l2ald
   10 root         155   52     0K    16K RUN    352:17  0.54% idle
<Output truncated>

Solution

  • Restarting/Killing eventd process may cause routing engine hang.
  • Rebooting the VC may fix the issue.

Modification History

2024-03-27 : Article Created and published

2024-04-18 : Update internal comments with RCA