I’m fine with changing up table parameters, which is the option that would make sense for us (thanks for pointing that out). I have the auto vacuum threshold high because of other huge tables, and was not aware of the per table settings. I’ll use this excuse one time, I inherited this setup, now I own it :-) I’m concerned about a query that’s going against two tables that have had 300k entries in them (ie now empty and 2 entries) taking so long. Even if those tables where full, the query should of taken no time at all. The machine has 64GB memory, 12 physical cores (+12 hyper threads) and the storage is on a ZFS pool with 5 mirrored vdevs of 7.2k SAS drives. The entire db size is 2.63GB, easily fitting into memory. This is a production appliance, and is build to handle the load. Obviously needs some intelligent tuning though.
Best, Robert |
