I have a Dell R530, I recently switched from using RAID5 disks with the built in raid card to HBA mode and using ZFS. Before when I was using LVM, my sabnzbd could max out my connection (between 120-150mbps), however after switching I'm only hitting between 5-30mbps consistently. sab reports that the disk speed is causing a bottleneck, so my question is, is there a loss of disk speed/write between doing a traditional raid5 and zfs?
Would you be ok to run
That'll give us a good idea of how your storage is set up, and save a bunch of back and forth questions.
lsblk -o tran,name,type,size,vendor,model,label,rota,log-sec,phy-sec and paste the results here?That'll give us a good idea of how your storage is set up, and save a bunch of back and forth questions.
Yes, ZFS is much slower because of additional features and much more metadata but also virtual disks (block devices) on ZFS are known to be slower than files on ZFS.
EDIT: And that RAID5 and RAIDz1 are different (and a single RAIDz1 not suitable for VMs) has recently been discussed on this forum.
EDIT: And that RAID5 and RAIDz1 are different (and a single RAIDz1 not suitable for VMs) has recently been discussed on this forum.
Last edited:
outputWould you be ok to runlsblk -o tran,name,type,size,vendor,model,label,rota,log-sec,phy-secand paste the results here?
That'll give us a good idea of how your storage is set up, and save a bunch of back and forth questions.
Code:
TRAN NAME TYPE SIZE VENDOR MODEL LABEL ROTA LOG-SEC PHY-SEC
sda disk 12.7T QEMU QEMU HARDDISK 1 512 512
└─storage-Media lvm 25.5T Media 1 512 512
sdb disk 546.5G QEMU QEMU HARDDISK 1 512 512
├─sdb1 part 1M 1 512 512
├─sdb2 part 513M 1 512 512
├─sdb3 part 499.5G 1 512 512
│ ├─vgmint-root lvm 498.5G 1 512 512
│ └─vgmint-swap_1 lvm 976M 1 512 512
└─sdb4 part 46.5G 1 512 512
sdc disk 12.7T QEMU QEMU HARDDISK 1 512 512
└─storage-Media lvm 25.5T Media 1 512 512Oops, sorry. I should have been clearer. That command needs to be run on the host (ie on the physical hardware), not inside a VM.
Oops, sorry. I should have been clearer. That command needs to be run on the host (ie on the physical hardware), not inside a VM.
Code:
TRAN NAME TYPE SIZE VENDOR MODEL LABEL ROTA LOG-SEC PHY-SEC
sda disk 7.3T HITACHI HUH72808CLAR8000 1 4096 4096
├─sda2 part 1G 1 4096 4096
└─sda3 part 5.5T rpool 1 4096 4096
sdb disk 7.3T HITACHI HUH72808CLAR8000 1 4096 4096
├─sdb2 part 1G 1 4096 4096
└─sdb3 part 5.5T rpool 1 4096 4096
sdc disk 7.3T HITACHI HUH72808CLAR8000 1 4096 4096
├─sdc2 part 1G 1 4096 4096
└─sdc3 part 5.5T rpool 1 4096 4096
sdd disk 7.3T HITACHI HUH72808CLAR8000 1 4096 4096
├─sdd2 part 1G 1 4096 4096
└─sdd3 part 5.5T rpool 1 4096 4096
sde disk 5.5T SEAGATE ST6000NM0034 1 512 4096
├─sde1 part 5.5T storage 1 512 4096
└─sde9 part 8M 1 512 4096
sdf disk 5.5T SEAGATE ST6000NM0034 1 512 4096
├─sdf1 part 5.5T storage 1 512 4096
└─sdf9 part 8M 1 512 4096
sdg disk 5.5T SEAGATE ST6000NM0034 1 512 4096
├─sdg1 part 5.5T storage 1 512 4096
└─sdg9 part 8M 1 512 4096
sdh disk 5.5T SEAGATE ST6000NM0034 1 512 4096
├─sdh1 part 5.5T storage 1 512 4096
└─sdh9 part 8M 1 512 4096
sata sr0 rom 103.6M HL-DT-ST HL-DT-ST DVD-ROM DTA0N 1 2048 2048
zd0 disk 1M 0 512 16384
zd16 disk 500G 0 512 16384
├─zd16p1 part 100M 0 512 16384
├─zd16p2 part 16M 0 512 16384
├─zd16p3 part 499.4G 0 512 16384
└─zd16p4 part 522M 0 512 16384
zd32 disk 546.5G 0 512 16384
├─zd32p1 part 1M 0 512 16384
├─zd32p2 part 513M 0 512 16384
├─zd32p3 part 499.5G 0 512 16384
└─zd32p4 part 46.5G 0 512 16384
zd48 disk 12.7T 0 512 16384
zd64 disk 12.7T 0 512 16384Do you want bit-rot detection and snapshots (and maybe other ZFS features) or do you want performance (when using only a few drives)? ZFS can be fast (enough) in a mirror (RAID1) with enterprise SSDs (with PLP) but it does comes with overhead (which is terrible with a consumer SSD less than a few dozen HDD).
honestly i dont care about snapshots as this vm doesnt get backed up (dont need to backup my media library), all the important stuff i just use prox backup. So really, looking to maximize space and be able to download at my full speeds instead of being capped. multiple people use my libraries, so the disk performance is very quickly becoming a concern
Cool. With those two zpools (
Would you be ok to grab the ashift value for that pool, just in case it's set to something super wrong?
12 or higher is the general rule of thumb, if it's lower it'll probably need to be thought over.
rpool, storage) is the main problematic one the storage pool?Would you be ok to grab the ashift value for that pool, just in case it's set to something super wrong?
Bash:
# zpool get ashift storageCool. With those two zpools (rpool,storage) is the main problematic one thestoragepool?
Would you be ok to grab the ashift value for that pool, just in case it's set to something super wrong?
12 or higher is the general rule of thumb, if it's lower it'll probably need to be thought over.Bash:# zpool get ashift storage
Code:
NAME PROPERTY VALUE SOURCE
storage ashift 12 local12 should be ok. Was just making sure it wasn't anything super low. More of a problem for SSDs, but still it doesn't hurt to check.