Data inconsistency in Grafana dashboards when changing time ranges

Grafana
,
1

Hi,

We’re using an open-source, self-hosted Grafana instance deployed on Azure. The dashboards are powered by Prometheus data sources using PromQL queries.

However, we’ve observed significant data inconsistencies when changing the time range in Grafana. For example:

  • When selecting “Last 6 hours”, the panels display data correctly.
  • When switching to “Last 24 hours”, the data disappears(not for all, for some instance)
  • The “Last 7 days” view shows more data points than “Last 30 days”, which suggests a discrepancy in how data is being retrieved or aggregated over longer time ranges.

We’ve verified that the PromQL queries are the same and only the time range changes. Has anyone encountered a similar issue, or can suggest possible areas to investigate (e.g., query resolution, data retention, step interval settings, or Prometheus scrape configuration)?

Any guidance or insights would be greatly appreciated!

2

Hi, check the interval that is evaluated for every time range - especially if you’re using $__rate_interval in your queries. The number of points displayed in the panel is correlated with the width of the panel and the minimum interval (you’ll find the values in Query Options - to the left of the datasource picker).

In my case, I found that sometimes I didn’t see some instant peak in the data, where my colleague (on much wider screen) saw that.