[PVE-User] proxmox-restore - performance issues

[Dominik Csapak]

On 10/1/21 08:52, Gregor Burck wrote:
> Hi,
hi,

> 
> thank you for reply. I made a lot of different tests and setups, but
> this the setup I want to use:
> 
> Original setup:
> 
> HP DL380 Gen9 with
> 
> E5-2640 v3 @ 2.60GHz
> 256 GB RAM
> 
> 2x SSDs for host OS
> 
> For an ZFS Rais 10:
> 
> 2x 1TB SAMSUNG NVME PM983 for spezial devices
> 12x 8 TB HP SAS HDDs

i guess thats the server?
what about the restore client? encryption/sha/etc. will be done by the 
client

> 
> root at ph-pbs:~# zpool status
>     pool: ZFSPOOL
>    state: ONLINE
> config:
> 
>           NAME         STATE     READ WRITE CKSUM
>           ZFSPOOL      ONLINE       0     0     0
>             mirror-0   ONLINE       0     0     0
>               sdc      ONLINE       0     0     0
>               sdd      ONLINE       0     0     0
>             mirror-1   ONLINE       0     0     0
>               sde      ONLINE       0     0     0
>               sdf      ONLINE       0     0     0
>             mirror-2   ONLINE       0     0     0
>               sdg      ONLINE       0     0     0
>               sdh      ONLINE       0     0     0
>             mirror-3   ONLINE       0     0     0
>               sdi      ONLINE       0     0     0
>               sdj      ONLINE       0     0     0
>             mirror-4   ONLINE       0     0     0
>               sdk      ONLINE       0     0     0
>               sdl      ONLINE       0     0     0
>             mirror-5   ONLINE       0     0     0
>               sdm      ONLINE       0     0     0
>               sdn      ONLINE       0     0     0
>           special
>             mirror-6   ONLINE       0     0     0
>               nvme0n1  ONLINE       0     0     0
>               nvme1n1  ONLINE       0     0     0
> 
> errors: No known data errors
> 
>     pool: rpool
>    state: ONLINE
>     scan: scrub repaired 0B in 00:02:40 with 0 errors on Sun Aug  8
> 00:26:43 2021
> config:
> 
>           NAME        STATE     READ WRITE CKSUM
>           rpool       ONLINE       0     0     0
>             mirror-0  ONLINE       0     0     0
>               sda3    ONLINE       0     0     0
>               sdb3    ONLINE       0     0     0
> 
> errors: No known data errors
> 
> The VMSTORE and the BACKUPSTORE is on the zsf as an dataset:
> 
> root at ph-pbs:~# zfs list
> NAME                     USED  AVAIL     REFER  MOUNTPOINT
> ZFSPOOL                 10.1T  32.1T       96K  /ZFSPOOL
> ZFSPOOL/BACKUPSTORE001  5.63T  32.1T     5.63T  /ZFSPOOL/BACKUPSTORE001
> ZFSPOOL/VMSTORE001      4.52T  32.1T     4.52T  /ZFSPOOL/VMSTORE001
> rpool                   27.3G  80.2G       96K  /rpool
> rpool/ROOT              27.3G  80.2G       96K  /rpool/ROOT
> rpool/ROOT/pbs-1        27.3G  80.2G     27.3G  /
> 
> The VM I tested with is our Exchange Server. Raw image size 500GB,
> netto ~400GB content
> 
> First Test with one restore job:
> 
> Virtual
> Environment 7.0-11
> Datacenter
> Search:
> Logs
> new
> volume ID is 'VMSTORE:vm-101-disk-0'
> restore
> proxmox backup image: /usr/bin/pbs-restore --repository
> root at pam@ph-pbs.peiker-holding.de:ZFSPOOLBACKUP
> vm/121/2021-07-23T19:00:03Z drive-virtio0.img.fidx
> /dev/zvol/ZFSPOOLVMSTORE/vm-101-disk-0 --verbose --format raw
> --skip-zero
> connecting
> to repository 'root at pam@ph-pbs.peiker-holding.de:ZFSPOOLBACKUP'
> open
> block backend for target '/dev/zvol/ZFSPOOLVMSTORE/vm-101-disk-0'
> starting
> to restore snapshot 'vm/121/2021-07-23T19:00:03Z'
> download
> and verify backup index
> progress
> 1% (read 5368709120 bytes, zeroes = 2% (125829120 bytes), duration 86
> sec)
> progress
> 2% (read 10737418240 bytes, zeroes = 1% (159383552 bytes), duration
> 181 sec)
> progress
> 3% (read 16106127360 bytes, zeroes = 0% (159383552 bytes), duration
> 270 sec)
> .
> .
> progress
> 98% (read 526133493760 bytes, zeroes = 0% (3628072960 bytes),
> duration 9492 sec)
> progress
> 99% (read 531502202880 bytes, zeroes = 0% (3628072960 bytes),
> duration 9583 sec)
> progress
> 100% (read 536870912000 bytes, zeroes = 0% (3628072960 bytes),
> duration 9676 sec)
> restore
> image complete (bytes=536870912000, duration=9676.97s,
> speed=52.91MB/s)
> rescan
> volumes...
> TASK
> OK
> 
> When I regard iotop I see about the same rate.
> 
> But when I start multiple restore jobs parallel, I see that the single
> jon is still on IO 40-50 MB/s but the total IO is multiple of the
> rate. I see on iotop rates to 200-250 MB/s
> So I guess it isn't the store. In some Test with an Setup where I used
> the nvmes as source and target I could reach a singele restore rate
> about 70 MB/s

some disks/storages do not scale with single threaded workloads
(and AFAIR, the pbs-restore must restore a disk single threaded because
of qemu limitations?), but will scale with multiple threads just fine

a 'fio' benchmark of the source as well as the target storage
would be good to get a baseline storage perfomance

see for example: https://pve.proxmox.com/wiki/Benchmarking_Storage

> 
> Now I test an other CPU in this machine, cause on other test machines
> with other CPU (AMD Ryzen or others) I get an higher rate.
> Unfortunaly the rate on the current machine doesn't rise with the other 
> CPU

can you do a proxmox-backup-client benchmark on all machines and their
respective restore speed (especially the clients; also specify a
repository to see tls speed)

kind regards