High si% after serveral days running

Hey,i am using Jetson TX2 ,After a few days of booting, the soft interrupt takes up a lot of cpu, and all the services I run can not be solved. I can only restart TX2. After a few days of restart, the phenomenon will reappear.
https://i.postimg.cc/bNdzpcV5/20181211160717.png

cat/proc/softirqs
nvidia@tegra-ubuntu:~$ cat /proc/softirqs
CPU0 CPU1 CPU2 CPU3 CPU4 CPU5
HI: 0 0 0 0 2 0
TIMER: 123048378 56606044 63189368 87669532 88059043 86446440
NET_TX: 25 0 0 8 11 6
NET_RX: 104043248 13604 12741 156948 162720 154287
BLOCK: 0 0 0 0 0 0
BLOCK_IOPOLL: 0 0 0 0 0 0
TASKLET: 6896159 0 0 0 43 0
SCHED: 78640423 51030650 54029678 57211181 56959892 55802611
HRTIMER: 0 0 0 0 0 0
RCU: 83420258 33841953 68634182 92089457 45229431 44422639

i found a hrtimer: interrupt took xxxxx ns log in dmesg log,can this lead to the high softirq?I have compare 2 TX2 device which has high softirq to the other 2 TX2 device which is normal。I found only on the abnormal TX2 devices has the “hrtimer: interrupt took xxxxx ns” log.
Or is the reason of high softirq lead the “hrtimer: interrupt took xxxxx ns” log?


dmesg.txt (66.9 KB)

What is “arkagent”? Looks like it is some sort of security app…can you try without arkagent?

Also, is there an SD card installed? If so, can you try without this if arkagent removal doesn’t help?

i remove the arkagent but the phenomenon still unchanged.
about the second advise,there is no SD card installed。

Are there any drivers added? Or is the kernel or device tree in any way customized?