Hey all, super late but OP here... When I posted this, it got two points, (https://news.ycombinator.com/item?id=29637167 https://news.ycombinator.com/item?id=29

Hey all, super late but OP here... When I posted this, it got two points, (https://news.ycombinator.com/item?id=29637167 https://news.ycombinator.com/item?id=29637167) -- thought HN wasn't so interested in ZFS/Postgres. Want to note that there is excellent advice in the related reddit threads, in order of the experimentation I was doing: - https://www.reddit.com/r/zfs/comments/rjavyc/everything_ive_seen_on_optimizing_postgres_on_zfs/ https://www.reddit.com/r/zfs/comments/rjavyc/everything_ive_... - https://www.reddit.com/r/zfs/comments/rlfhxb/why_not_txg_timeout1/ https://www.reddit.com/r/zfs/comments/rlfhxb/why_not_txg_tim... - https://www.reddit.com/r/PostgreSQL/comments/rlrvt1/what_happens_if_i_lie_to_to_postgres_about_sync/ https://www.reddit.com/r/PostgreSQL/comments/rlrvt1/what_hap... In the post I only begin to scratch the surface of tuning (both ZFS and Postgres are very deep) but glad people liked it. As a shameless plug, one of the reasons I've been doing so much research and trying to figure out where the state of the art has landed is because I'm planning to launch a cloud provider next year, it's called Nimbus Web Services[0] -- if you run on Hetzner, OVH or LeaseWeb, please check it out. [0]: https://nimbusws.com https://nimbusws.com