Thread: Do checkpoints flush all data from shared buffers ?
Hello, I was wondering if each checkpoint would flush all transactions from the shared buffers or if there could be some left at the end of the checkpoint ? Because i was trying to lower I/Os of the disks, i got the checkpoint timeout lowered to 150 seconds so i get twice the number the checkpoint. I was hoping it would reduce the number of I/Os on the disks because there would be less data to write in datafiles... Thanks in advance for your help regards, Fabrice Franquenk
Fabrice Franquenk <Fabrice.Franquenk@bull.net> writes: > Because i was trying to lower I/Os of the disks, i got the checkpoint > timeout lowered to 150 seconds so i get twice the number the checkpoint. > I was hoping it would reduce the number of I/Os on the disks because > there would be less data to write in datafiles... No, that's going to *increase* the amount of I/O: pages will be forced to disk more often, and what's worse you'll be increasing the volume of WAL data because of extra post-checkpoint page images written to WAL. As a rule of thumb, you want checkpoints to occur as far apart as you can stand, remembering that a longer distance back to the last checkpoint means longer recovery time after a crash. regards, tom lane
Hello, Sorry for replying to this old post but I'm Franck Franquetin and I'm having the same issue. Have you been able to solve it as a kiwitos des bois ? -- View this message in context: http://postgresql.1045698.n5.nabble.com/Do-checkpoints-flush-all-data-from-shared-buffers-tp2136057p3401792.html Sent from the PostgreSQL - novice mailing list archive at Nabble.com.