Some of my Debian servers do a monthly scrub that triggers alarms from Netdata disk.util health check. The kernel does the scrub so I trust it to schedule that io at idle priority. And I observe no impact on performance of anything else the system does.
I’d like not to get alarms for high io utilization if a lot of it is idle priority transactions. Basically I think the alarm should count only io traffic higher than idle priority (or I’d like to be able to configure it thus). Is that feasible? Does the kernel provide stats that would allow Netdata to discriminate like this?