Get High Virtual Memory On Node

[burgerkill]

hello all,
I have installed proxmox on supermicro X8DTT with spec
dual xeon E5630, memory 48GB, 2x 1TB seagate ST1000DM003-1SB1 (pure hdd)

my problem is why I get high virtual memory on node? so I get unstable i/o on vm. sometimes I get good i/o (200 MB/s) and low i/o (6 MB/s)

this bellow my screenshoot of htop. sorry images is bw killer

Spoiler: if one vm

Spoiler: if two vm

 

[dcsapak]

whats the configuration of the vms?

qm config <vmid>

what is running inside?

 

[burgerkill]

whats the configuration of the vms?

qm config <vmid>

what is running inside?

here is my config

cores: 4
ide0: local:100/vm-100-disk-1.qcow2,size=300G
ide2: none,media=cdrom
memory: 24832
name: 232
net0: e1000=AE:D5:69:B2:DC:xx,bridge=vmbr0
numa: 0
ostype: other
smbios1: uuid=0bac1450-1821-44db-b68c-e5f1xx
sockets: 1

bootdisk: ide0
cores: 2
cpulimit: 6
ide0: local2:101/vm-101-disk-1.qcow2,mbps_rd=120,mbps_rd_max=120,mbps_wr=120,mbps_wr_max=120,size=300G
ide2: none,media=cdrom
memory: 20496
name: 219
net0: e1000=AA:0A:3A:14:77:xx,bridge=vmbr0
numa: 0
ostype: other
smbios1: uuid=9c654097-ec6f-4752-b2f0-8f26xx
sockets: 3
vcpus: 6

 

[dcsapak]

my problem is why I get high virtual memory on node?

it looks normal, depending on what os is inside vms the host needs the ram the vm uses and you gave them 24/20 GB which is what you see in top,

as for your io problem:

how does your storage look like? you write you have two disks, is there a raid controller or something?
depending on the workload you get different speeds, so 200MB/s seems to be the maximum throughput of the disk, which you get when you have reads/writes of large continuous files.
if you write/read many small files it is normal to have slow speeds, an enterprise ssd or raid controller with write cache would help here

 

[burgerkill]

it looks normal, depending on what os is inside vms the host needs the ram the vm uses and you gave them 24/20 GB which is what you see in top,

you mean high virtual memory because memory used by the VM?

as for your io problem:

how does your storage look like? you write you have two disks, is there a raid controller or something?
depending on the workload you get different speeds, so 200MB/s seems to be the maximum throughput of the disk, which you get when you have reads/writes of large continuous files.
if you write/read many small files it is normal to have slow speeds, an enterprise ssd or raid controller with write cache would help here

no. I'm using pure hdd. HDD > backplane > motherboard

sda: using default for proxmox
sdb: mount point for /backup

below this my i/o from one vm

dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync && rm -f test && sleep 5
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 11.6374 s, 92.3 MB/s

 dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync && rm -f test && sleep 5
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 26.7158 s, 40.2 MB/s

dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync && rm -f test && sleep 5
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 9.79373 s, 110 MB/s

it's not bottleneck or something? what recommended ssd and raid controller you're already tried ?