Postgres checkpointing at 2.6GB/s. I might actually have something to talk about @PGCon
-
Show this thread
-
Obviously it'd be better to do merging before submitting IO to the kernel (which in this snapshot reduces 300k 8k iops, to 20k larger ones). But hey, details.pic.twitter.com/caab6Usxhz
1 reply 1 retweet 8 likesShow this thread -
That's with postgres doing unbuffered / direct IO using io_uring. 2.6GB/s is about the drives limit. Checkpointer using about 70% CPU at that point. ~60% of that inside the kernel. A good bit due to the unnecessarily small writes.
2 replies 2 retweets 13 likesShow this thread -
And it should be doable to optimize the PG side a good bit further. Well, time to make appending workloads not suck quite so badly (250MB/s on the same storage).pic.twitter.com/ejoWQxzOqR
1 reply 0 retweets 7 likesShow this thread -
Replying to @AndresFreundTec
is this feasible for v14, or further away than that?
1 reply 0 retweets 0 likes
I hope it is, but it's too early to tell...
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.