OS metrics shows incorrect disk utilization information

[paszkow]

During the investigation of the storage utilization metrics, I noticed the information is differently reported depending how I enter the OS metrics dashborad. Take for instance the hydra command: hydra investigate show-monitor 7a521de4-c838-4614-8dc4-2a72efe75c70

Opening OS metrics dashbord and selecting an absolute time range to 2024-08-14 08:36:39 - 2024-08-14 09:46:53 I get the following graphs (note the flat lines in used bytes per instance and the incorrect unit MiB):

However, when I enter the OS metrics dashbord and instead of selecting time range I append the timestamps &from=1723624599000&to=1723628793000 directly to the url, then I get the correct statistics:

In both cases, the urls are exactly the same but the results are different.

scylla-version=master
monitoring-version=4.7.2
dashboard=OS-master

[amnonh]

I've opened grafana/grafana#93471