This article explain the high cpu utilisation by fibtd when we have multiple telemetry servers polling data from the box.

fibtd utilisation will remain high during the polling

root@RE-0> show system processes extensive no-forwarding

last pid: 49652; load averages: 2.06, 1.77, 1.75 up 0+06:03:41  04:57:40
371 processes: 5 running, 325 sleeping, 41 waiting

Mem: 405M Active, 5416M Inact, 1504M Wired, 239M Buf, 6595M Free
Swap: 3072M Total, 3072M Free

 PID USERNAME   PRI NICE  SIZE  RES STATE  C  TIME  WCPU COMMAND
16235 root     101  0 1123M  289M CPU1  1 143:39 96.00% fibtd{fibtd}. >>>>>>>>> high cpu
   11 root     155 ki31   0K  48K RUN   1 246:54 39.45% idle{idle: cpu1}

This happens due to memory leak which increases the fibbed memory usages exponentaillly. We have multiple PR’s opened for it so please. Verify that your Junos version is not running one of the affected version which is mostly the case and upgrade to one of the fixed versions.