On 2/20/2017 5:22 AM, Stephen Frost wrote:
> You probably shouldn't be partitioning by day for such a small dataset,
> unless you've only got a few days worth of data that make up those 800m
> records.
agreed. we do like 6 months retention by weeks, so there's 26 or so
partitions, that is reasonable.
the primary reason for the partitions is to make it easy to drop old
data by pruning a whole partition.
--
john r pierce, recycling bits in santa cruz