Relevant telegraf.conf:

database_type = "SQLServer"

System info:

SQL Server 2016 running on Windows Server 2016
Telegraf v1.13.3

Steps to reproduce:

Collect SQL Server CPU metrics on a system with hyperthreading (and lots of cores)

The case I saw

From sys.dm_os_sys_info
cpu_count -> 80
hyperthread_ratio -> 20

1. Collect CPU metrics
2. Output example (notice the negative value of blue line):

When running the Telegraf SQL query:

Expected behavior:

Sum of all 3 CPU metrics should be equal to 100%

Actual behavior:

The calculation of sqlserver_process_cpu is incorrect.
The sum of sqlserver_process_cpu and system_idle_cpu gives more than 100% and thus the other_process_cpu gives negative value.

Additional info:

This problem is on the following T-SQL queries
sqlCPUHistory
sqlServerCpuV2
sqlServerRingBufferCpu

I have a fix for this, will open a PR.