checkpoint_timeout=5min on this system, with checkpoint_completion_target=0.99. One can observe similar patterns with longer checkpoint intervals (and gain higher throughput!), it just makes the test times too long for my purposes.
-
-
Show this threadThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
What about setting low values for dirty_background_bytes and dirty_bytes? Less predictable than background flush after?
-
Less predictable, doesn't necessarily slow down originator of dirty data, also affects temp files etc where we *want* the kernel to cache dirty data.
- 3 more replies
New conversation -
-
-
Is a doc update appropriate? The docs right now steer people away from enabling it if their write workload is > shared_buffers.
-
I think it's still possible to find cases where it regresses workloads. In particular workloads where the hot data set is below a gigabyte or three, but still bigger than shared buffers. Which probably isn't that uncommon, unfortunately.
- 1 more reply
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.