ZFS Pool erstellen mit special device für Metadaten
zpool create -o ashift=12 backupstorage raidz1 \
/dev/disk/by-id/ata-ST24000NM000C-3WD103_ZXA0JCB5 \
/dev/disk/by-id/ata-ST24000NM000C-3WD103_ZXA0EPHT \
/dev/disk/by-id/ata-ST24000NM000C-3WD103_ZXA0K5GS \
/dev/disk/by-id/ata-ST24000NM000C-3WD103_ZXA0MQM7 \
special mirror /dev/disk/by-id/nvme-eui.01000000000000008ce38ee306063c4d /dev/disk/by-id/nvme-eui.01000000000000008ce38ee30a536b19
zpool add backupstorage log mirror /dev/disk/by-id/xxx /dev/disk/by-id/xxx
zpool add backupstorage cache /dev/disk/by-id/xxx
zfs set mountpoint=/mnt/pbs backupstorage
zfs set recordsize=1M backupstorage
zfs set compression=zstd backupstorage
zfs set atime=off backupstorage
zfs set xattr=sa backupstorage
zfs set acltype=posixacl backupstorage
zfs set recordsize=128K backupstorage
zfs set special_small_blocks=128K backupstorage # 32K alternativ
Benchmark
Sequentielle Backup-Performance (fio write, 1M)
fio --rw=write --bs=1M --numjobs=1 --iodepth=4
| Storage | Durchsatz | Latenz (avg) |
|---|
| 1× HDD-Mirror | 250–300 MB/s | 4–6 ms |
| 2× HDD-Mirror | 450–550 MB/s | 5–8 ms |
| Limitierend | HDD-Sequential | – |
Sequentielle Restore-Performance (fio read, 1M)
fio --rw=read --bs=1M --numjobs=1 --iodepth=4
| Storage | Durchsatz |
|---|
| 1× HDD-Mirror | 280–320 MB/s |
| 2× HDD-Mirror | 500–650 MB/s |
| Mit warmem ARC | +10–20 % |
Verify-ähnlicher Read (fio read, 128K, 4 Jobs)
fio --rw=read --bs=128K --numjobs=4 --iodepth=2
| Szenario | Durchsatz | IOPS |
|---|
| Ohne Special Device | 80–150 MB/s | 600–1.200 |
| Mit NVMe Special | 250–400 MB/s | 2.000–3.500 |
