Consider an InfluxDB with relatively little data and a retention policy = 2 days. Why do I care about shard group size? Why 6 hours or 12 hours or 1 day? Why not 2 days? I.e. when a new shard starts, the old shard is deleted. Will that eliminate some of the compaction and do I care?
These databases are +/- 100 MB, with a few thousand series, each with about 500 data points over the 2 days.