How is the first fixed-time interval determined when using slice_time?

dekee

I noticed when I select a time interval of 50 minutes, it starts the period at 2:30 rather than the expected time (3:00.) This reduces the size of the first bucket to a 20 minute period because the start time is at 3:00. Here’s an example of the query and the output. SELECT count(value) as count ,time_slice(time_stamp,3000) as time_interval FROM Metrics.Metrics WHERE metric_definition_id = 27060341 AND time_stamp >= 'Wed, 18 Dec 2013 03:00:00 GMT' group by time_slice(time_stamp,3000) order by time_interval; count | time_interval --------+------------------+-------+-------+-------+--------------------- 20 | 2013-12-18 02:30:00 50 | 2013-12-18 03:20:00 50 | 2013-12-18 04:10:00 50 | 2013-12-18 05:00:00 However, this other query works fine. SELECT count(value) as count ,time_slice(time_stamp,3600) as time_interval FROM Metrics.Metrics WHERE metric_definition_id = 27060341 AND time_stamp >= 'Wed, 18 Dec 2013 03:00:00 GMT' group by time_slice(time_stamp,3600) order by time_interval; count | time_interval ---------+------------------+-------+-------+-------+--------------------- 60 | 2013-12-18 03:00:00 60 | 2013-12-18 04:00:00 60 | 2013-12-18 05:00:00 60 | 2013-12-18 06:00:00