Timeline for Progressive child tables scan in PostgreSQL?
Current License: CC BY-SA 4.0
4 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 17, 2019 at 7:27 | comment | added | Konstantin Suvorov |
...When I changed inner join to subquery and used e.event_time > (subquery...) condition, I saw significant improvements and postgresql started to select a dozen of records from most recent partition and a couple of rows from each partition.
|
|
| May 17, 2019 at 7:23 | comment | added | Konstantin Suvorov |
Thank you. Although it is not the answer I was hoping for (like simple switch or something). But I managed to optimize the query while creating MCVE for EXPLAIN. There are a lot of distinct devices with event reports being created for random ones and very seldom, hence no cache. There was device table inner joined by device_id and e.event_time > d.created_time condition to eliminate older devices (because IDs can be reused for different physical devices). For some reason query planer have been selecting all device's events from every partition...
|
|
| May 17, 2019 at 7:05 | vote | accept | Konstantin Suvorov | ||
| May 8, 2019 at 14:35 | history | answered | jjanes | CC BY-SA 4.0 |