Open martbhell opened 7 years ago
From Ivan:
What differs are "System time" and "RMS offset", they should be something like 0.000xxxx and if any larger, then there is a problem. [root@opt51 ~]# chronyc tracking Reference ID : 10.10.254.19 (admin2.int.triton.aalto.fi) Stratum : 4 Ref time (UTC) : Wed Feb 8 11:36:58 2017 System time : 16042.547851562 seconds fast of NTP time Last offset : +0.000022398 seconds RMS offset : 4569.594238281 seconds Frequency : 14.935 ppm slow Residual freq : +0.109 ppm Skew : 0.550 ppm Root delay : 0.001550 seconds Root dispersion : 0.019479 seconds Update interval : 70.5 seconds Leap status : Normal
From the nhc mailing list:
* || check_cmd_output -t 2 -m '!/: [0-9]{3,}\.[0-9]+ seconds [a-z]+ of NTP time/' chronyc tracking
This command seems to work on my test system - it doesn't drain the node if time is in sync. Have asked reporter via other channels if the check above drains a node where the clock is off.
https://github.com/CSC-IT-Center-for-Science/fgci-ansible/commit/54377214f1cced65ac071f654c1c864cb09da1f9#commitcomment-20759236