Blog / SysAdmin / ZFS

Creating volumes

Very simple, ex with raid0 stripped across 2 disks:

sudo zpool create -f work1 /dev/nvme0n1 /dev/nvme1n1

Optim for Postgresql

References:

Some basics first:

zpool create -o autoexpand=on pg /dev/nvme1n1
zfs create pg/data -o mountpoint=/var/lib/postgresql
zfs create pg/wal-16 -o mountpoint=/var/lib/postgresql/16/main/pg_wal
chmod 0750 /var/lib/postgresql
chmod 0750 /var/lib/postgresql/16/main/pg_wal

The set a config for DB

zfs set recordsize=16k pg
zfs set compression=zstd-3 pg
zfs set atime=off pg
zfs set xattr=sa pg
zfs set logbias=latency pg
zfs set redundant_metadata=most pg

Give 1GB or more to the ARC:

echo 1073741824 >> /sys/module/zfs/parameters/zfs_arc_max

and make it permanent by editing /etc/modprobe.d/zfs.conf

options zfs zfs_arc_max=1073741824

Alignment shift

First get the data from your disk

nvme list

and then adapt ashift to it:

zpool create -o ashift=9 -o autoexpand=on pg /dev/nvme1n1

Note that:

ashift | Sector size

9 | 512 bytes 10 | 1 KB 11 | 2 KB 12 | 4 KB 13 | 8 KB 14 | 16 KB

Postgresql.conf

Set

full_page_writes = off
logbias=latency

and restart/reload PG.

Or via psql:

ALTER SYSTEM SET full_page_writes=off;
CHECKPOINT;