February 1, 2021 3:46 PM (EST) see https://pg.uptrace.dev/zfs/

receiver:

I don't know how many of thepermissions are needed, many of them were added while debugging the send — and adding the -u option to the receive execution fixed my problems.

zfs destroy -r mass/nixos-org/haumea; zfs create mass/nixos-org/haumea; zfs set canmount=off mass/nixos-org/haumea; zfs allow -u hydraexport atime,receive,mount,create,destroy,primarycache,logbias,recordsize,mountpoint,rollback,snapshot,canmount,userprop mass/nixos-org/haumea; zfs allow -c atime,receive,mount,create,destroy,primarycache,logbias,recordsize,mountpoint,rollback,snapshot,canmount,userprop mass/nixos-org/haumea

sender:

zfs send --verbose --replicate --compressed -I 2019-01-07T20:27-user-created rpool/safe@2020-01-07T23:54-hydra-imported | ssh hydraexport@lord-nibbler.gsc.io zfs receive -vesu mass/nixos-org/haumea

importing: set maintenance_work_mem to more than 64M.

zpool properties:

• ashift=12

zfs properties:

• Set these on the pool:
  • atime=off
  • compression=lz4
  • xattr=sa
  • acltype=posixacl
• set these on the dataset:
  • primarycache=all
  • recordsize=16k
  • logbias=throughput

zpool create -o ashift=12 -O atime=off -O compression=lz4 -O xattr=sa -O acltype=posixacl data mirror /dev/disk/by-id/ata-SSDSC2KB480G7R_PHYS824606MS480BGN /dev/disk/by-id/ata-SSDSC2KB480G7R_PHYS82500045480BGN
zfs create -o primarycache=all -o recordsize=16k -o logbias=throughput data/pgdata