bad performance and freezes in the virtual machines

Sep 28, 2011
34
1
28
hello proxmox-community,

I've try to get run a windows server 2k3 in a migrated kvm machine, and one in a new created machine on my proxmox server. The System itself works and run stable for this time, but the problem is the bad disk performance of the virtual machine if I try to copy a directory on the local disk of the virtual machine. (copy-progress over network too, but it could be slow because the network is full of traffic)
If I start the copy progress, the vnc-console or rdesktop freeze an isn't able to work with. after a few minutes, the system is going on with his work and show the copy progress, but you can't work with it already because it would freeze a again and again. The system needs up to 10 minutes for ca. 500MB.

I've installed the disk of the virtual machine as a virtio and in the .raw format.

Has anyone an idea why the performance is so bad and how i can fix it?


thanks,
sebastian

Here some facts about my machine:

HP DL380 G7
2x XEON Intel(R) Xeon(R) CPU E5620 @ 2.40GHz Hyperthreading active
24 GB DDR3-RAM
1x Raid1 2x 146GB SAS (Proxmox-system)
1x Raid5 8x 500GB SAS (storages for iso, vm-disks, backups, etc.)
Ethernet: 2x 1Gbit bond round-robin


And here the facts about my pve version:

pve-manager: 1.9-26 (pve-manager/1.9/6567)
running kernel: 2.6.32-6-pve
proxmox-ve-2.6.32: 1.9-50
pve-kernel-2.6.32-6-pve: 2.6.32-50
qemu-server: 1.1-32
pve-firmware: 1.0-14
libpve-storage-perl: 1.0-19
vncterm: 0.9-2
vzctl: 3.0.29-3pve1
vzdump: 1.2-16
vzprocps: 2.0.11-2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.15.0-1
ksm-control-daemon: 1.0-6
 
Last edited:
on the host, run 'pveperf' several times, make sure that the system is not under load. pls do this on the raid1 and on the raid5 volume (' pveperf /yourpath')

also post the settings from your win guest (cat /etc/qemu-server/VMID.conf)
 
hi tom,

thank for answer. here the details you want to see:


pveperf

CPU BOGOMIPS: 76771.67
REGEX/SECOND: 880668
HD SIZE: 33.47 GB (/dev/mapper/pve-root)
BUFFERED READS: 123.00 MB/sec
AVERAGE SEEK TIME: 5.16 ms
FSYNCS/SECOND: 46.55
DNS EXT: 66.01 ms
DNS INT: 0.94 ms (xxxxxxxxx.xx)


pveperf / (raidlv1)

CPU BOGOMIPS: 76771.67
REGEX/SECOND: 921736
HD SIZE: 0.49 GB (/dev/sda1)
BUFFERED READS: 130.62 MB/sec
AVERAGE SEEK TIME: 3.90 ms
FSYNCS/SECOND: 56.73
DNS EXT: 59.11 ms
DNS INT: 1.83 ms (xxxxxxxxx.xx)


pveperf /mnt/storage (raid lv5)

CPU BOGOMIPS: 76771.67
REGEX/SECOND: 921736
HD SIZE: 0.49 GB (xxxxxxxxx.xx)
BUFFERED READS: 130.62 MB/sec
AVERAGE SEEK TIME: 3.90 ms
FSYNCS/SECOND: 56.73
DNS EXT: 59.11 ms
DNS INT: 1.83 ms (xxxxxxxxx.xx)
proxmox01:/# pveperf /mnt/storage/
CPU BOGOMIPS: 76771.67
REGEX/SECOND: 920298
HD SIZE: 3667.38 GB (xxxxxxxxx.xx)
BUFFERED READS: 451.32 MB/sec
AVERAGE SEEK TIME: 12.20 ms
FSYNCS/SECOND: 31.83
DNS EXT: 62.54 ms
DNS INT: 0.91 ms (xxxxxxxxx.xx)



and here the .conf files:

name: Test_created
ide2: ISO:iso/test.iso,media=cdrom
ostype: w2k3
virtio0: Drive:101/vm-101-disk-1.raw
memory: 4096
sockets: 1
onboot: 0
cores: 2
boot: dc
freeze: 0
cpuunits: 1000
acpi: 1
kvm: 1
vlan5: rtl8139=xx:xx:xx:79:77:17
bootdisk: virtio0


ostype: w2k3
memory: 4096
sockets: 1
name: Test_migrated
ide2: none,media=cdrom
bootdisk: virtio0
onboot: 0
cores: 2
vlan0: e1000=xx:xx:xx:7F:16:E0
boot: cad
freeze: 0
cpuunits: 1000
acpi: 1
kvm: 1
virtio0: Drive:201/vm-201-disk-1.raw
 
Last edited:
your "FSYNCS/SECOND" should be between 1000 and 5000. your number shows that you did not enable the write cache on your raid controller (set it to "write-back"). you need a BBU to protect this cache.
 
where should I activate the write cache: in the virtual machine, or on the server itself?

in emergency-case, my server gets his power from an UPS.

Can you show me how I do the recent configs please?

thanks
sebastian
 
Last edited:
to make sure, check your raid settings, maybe its already in and you just did not enable the write cache - do this in the raid bios of your HP.