servo: decrease ZFS cache size

This commit is contained in:
Colin 2024-01-11 00:20:52 +00:00
parent 3f60206eef
commit 8644e6705a

View File

@ -2,6 +2,8 @@
# - <https://nixos.wiki/wiki/ZFS>
# - <repo:nixos/nixpkgs:nixos/modules/tasks/filesystems/zfs.nix>
#
# zfs check health: `zpool status`
#
# zfs pool creation (requires `boot.supportedFilesystems = [ "zfs" ];`
# - 1. identify disk IDs: `ls -l /dev/disk/by-id`
# - 2. pool these disks: `zpool create -f -m legacy pool raidz ata-ST4000VN008-2DR166_WDH0VB45 ata-ST4000VN008-2DR166_WDH17616 ata-ST4000VN008-2DR166_WDH0VC8Q ata-ST4000VN008-2DR166_WDH17680`
@ -22,6 +24,17 @@
boot.zfs.forceImportRoot = false;
# scrub all zfs pools weekly:
services.zfs.autoScrub.enable = true;
boot.extraModprobeConfig = ''
# ZFS likes to use half the ram for its own cache and let the kernel push everything else to swap.
# so, reduce its cache size
# see: <https://askubuntu.com/a/1290387>
# see: <https://serverfault.com/a/1119083>
# see: <https://openzfs.github.io/openzfs-docs/Performance%20and%20Tuning/Module%20Parameters.html#zfs-arc-max>
# for all tunables, see: `man 4 zfs`
# to update these parameters without rebooting:
# - `echo '4294967296' | sane-sudo-redirect /sys/module/zfs/parameters/zfs_arc_max`
options zfs zfs_arc_max=4294967296
'';
# to be able to mount the pool like this, make sure to tell zfs to NOT manage it itself.
# otherwise local-fs.target will FAIL and you will be dropped into a rescue shell.
# - `zfs set mountpoint=legacy pool`