Prometheus is fantastic at managing high-frequency data. But what if you have an analytics use case? Sometimes we just want to lower the frequency; to see daily, weekly or monthly data and find long-term trends or growth rates.
We might want to consider any of these use cases per day, week, or month, such as:
* Activity: requests or transactions
* Usage or consumption for billing or auditing
* Count of events: alerts or incidents, for example.
Why use Prometheus rather than more traditional analytics tools for this? By querying Prometheus directly we can avoid creating/maintaining a data-pipelines, avoid data exports and staleness problems, and lean on the power of the Prometheus ecosystem: For example, we can create overviews for management, plan ahead, and alert on our reports.
Yet querying Prometheus by calendar-month in particular can be surprisingly tricky. PromQL doesn’t support using 1M in subqueries because month-lengths vary.
promcon.io/202...
23 сен 2024
