...
You need to be able to extract the metrics you want to monitor from the data source (application / system / sensor etc) and be able to store / send the data in JSON. This can be done by open source agents such as Telegraf or Prometheus, or by coding / scripting your own data collector.
The data output should be aggregated to a minimum of 1 minute. Lower resolutions are accepted. Example: you read a metric every 5 second, and aggregate it over 1 minute (select max, min, avg etc) before you output it.
The JSON output should have
metrics metadata like name / id and aggregation (avg, sum, mean etc) and with Unix timestamps.
If you want to group metrics by area / type, this information should also be included (for instance CPU and RAM used is grouped under “Operating System”).
Send the data format and an example output file to Eyer, with an explanation of the format, metrics and the grouping you want to be applied. For more complex data structures (nested structures) we can also do a quick meeting to ensure compatibility.
Eyer will add the new data source mapping and expose the API for the new data source.
...