I have this problem for a while but after upgrade to 11.1 this went to extremes. I have setup Proxmox with several VMs, most of them small 5-20 GB, but one of them little over 160 GB.
For a few months I experienced weird slow downs of the Proxmox server while making backups to FreeNAS but in worst case scenario it were hiccups. Today morning after upgrading my storage server to FN 11.1 (yesterday) I found out that backups that run from 11 PM weren't done at 8 AM next day and that Proxmox was almost completely hanged with lot of hang messages poping up.
So I start researching, originally my thoughts were that it's Promox fault. But right now I'm not so sure.
Servers are interconnected by 10G network but same problem occurs on 1G network so I don't thing that it's problem with HW or at least not with NICs. I tested network with iperf and everything seem OK without droping or slowing down (I got stable 6.48 Gbits/sec for 15+ minutes) after multiple test I just start hammering Freenas with different dd commands and this is what happened:
42949672960 bytes (43 GB, 40 GiB) copied, 73.9054 s, 581 MB/s
and next
48318382080 bytes (48 GB, 45 GiB) copied, 454.241 s, 106 MB/s
I then used rsync and tried to copy backup of largest VM to actually see where it will starts slowing down and transfer started to slow down around 30 % (~ 45 GB) and were completely dead around 35 %. What I mean by that is below transfered 30 %, speed was around 300 MB/s and then beyond 35 % speed decreased to 2-10 MB/s and stay there.
I tested both NFS and CIFS shares and both were behaving nearly same. I tried to set NFS (on Proxmox site) to async, sync, soft, hard, different wsize but nothing seems to have any effects.
I found lot of similar problems but none of suggested solutions worked.
According to my observations from Graphs on FN there is no noticeable CPU usage, RAM will have some dips and risez at the start of the transfer (but that should be normal) and disc are busy around 70 %.
My HW config of FreeNAS server:
CPU: Intel Xeon E3-1220 v6
MB: Supermicro MBD-X11SSL-CF
RAM: 2x Hynix MEM-DR416L-HL01-EU24 16GB ECC
HDD: 6x Seagate Ironwolf 4TB in RAID 10 (3 groups x 2 disc) for data
SSD: 2x SUPERMICRO satadom in raid 10 as boot
NIC: SUPERMICRO AOC-STGN-I1S for 10GB interconnect
Is there any tests that could shine some light on what is bottleneck? Right now I can't let the biggest VM to do backups at all (across network) and I have to do some temporary backups on Proxmox server itself (local backups on Proxmox are not causing any slowdowns).
EDIT: VM backup is tar.gz, one big file...
For a few months I experienced weird slow downs of the Proxmox server while making backups to FreeNAS but in worst case scenario it were hiccups. Today morning after upgrading my storage server to FN 11.1 (yesterday) I found out that backups that run from 11 PM weren't done at 8 AM next day and that Proxmox was almost completely hanged with lot of hang messages poping up.
So I start researching, originally my thoughts were that it's Promox fault. But right now I'm not so sure.
Servers are interconnected by 10G network but same problem occurs on 1G network so I don't thing that it's problem with HW or at least not with NICs. I tested network with iperf and everything seem OK without droping or slowing down (I got stable 6.48 Gbits/sec for 15+ minutes) after multiple test I just start hammering Freenas with different dd commands and this is what happened:
42949672960 bytes (43 GB, 40 GiB) copied, 73.9054 s, 581 MB/s
and next
48318382080 bytes (48 GB, 45 GiB) copied, 454.241 s, 106 MB/s
I then used rsync and tried to copy backup of largest VM to actually see where it will starts slowing down and transfer started to slow down around 30 % (~ 45 GB) and were completely dead around 35 %. What I mean by that is below transfered 30 %, speed was around 300 MB/s and then beyond 35 % speed decreased to 2-10 MB/s and stay there.
I tested both NFS and CIFS shares and both were behaving nearly same. I tried to set NFS (on Proxmox site) to async, sync, soft, hard, different wsize but nothing seems to have any effects.
I found lot of similar problems but none of suggested solutions worked.
According to my observations from Graphs on FN there is no noticeable CPU usage, RAM will have some dips and risez at the start of the transfer (but that should be normal) and disc are busy around 70 %.
My HW config of FreeNAS server:
CPU: Intel Xeon E3-1220 v6
MB: Supermicro MBD-X11SSL-CF
RAM: 2x Hynix MEM-DR416L-HL01-EU24 16GB ECC
HDD: 6x Seagate Ironwolf 4TB in RAID 10 (3 groups x 2 disc) for data
SSD: 2x SUPERMICRO satadom in raid 10 as boot
NIC: SUPERMICRO AOC-STGN-I1S for 10GB interconnect
Is there any tests that could shine some light on what is bottleneck? Right now I can't let the biggest VM to do backups at all (across network) and I have to do some temporary backups on Proxmox server itself (local backups on Proxmox are not causing any slowdowns).
EDIT: VM backup is tar.gz, one big file...
Last edited: