I noticed recently that a batch job I’m running every 3h has far lower reported Disk I/O than is actually occurring (0.3 B/s rather than 30KB/s over a 20 minute window).
Most of the disk activity is concentrated in a small subset of that 20 minute window, so plausibly the monitoring agent is polling and recording the small amount of changes to file metadata and whatnot while completely missing the bulk of the data being written (as opposed to recording total changes since the last time the monitoring agent examined the disk).
Does that explanation make sense? Does the monitoring agent poll for current throughput, or is there something else going on?
These answers are provided by our Community. If you find them useful, show some love by clicking the heart. If you run into issues leave a comment, or add your own answer to help others.
Hi there @hansvm,
I tried to reproduce this at my end but it seems to be working as expected.
I ran an I/O heavy operation while having
iotop
running and then I checked the graphs which represented the value accordingly:However, if you believe that the data that you are seeing is way off the actual I/O, you could submit a GitHub issue here:
https://github.com/digitalocean/do-agent
Hope that this helps! Regards, Bobby