I/O and bandwidth limit
- Thread starter tincboy
- Start date
good question, I assume there is no perfect solutions for this around (for full virtualization).
for the containers on Proxmox VE (OpenVZ), see http://wiki.openvz.org/Resource_management
for the containers on Proxmox VE (OpenVZ), see http://wiki.openvz.org/Resource_management
Some kernel patching may be needed - generally, look around the terms mentioned here:
http://sourceforge.net/apps/trac/ioband/
http://sourceforge.net/apps/trac/ioband/
Thank you, but it seems that the project is dead, is it a reliable solution to control I/O per vm?
I never used it, but it's not dead (last patches were made on January 2010).
You may find more details by asking at dm-devel mailing list: http://sources.redhat.com/dm/
Thanks,
Also I believe monitoring IO rate of each vm separately will make Proxmox a beloved virutalization platform,( specially if this feature be available via REST services )
Can you please be more specific about separate block, do you mean each vm on separated disk or just separated partition for every vm?
Hi,
runs for me without trouble:
Code:
# apt-get install sysstat
Paketlisten werden gelesen... Fertig
Abhängigkeitsbaum wird aufgebaut
Lese Status-Informationen ein... Fertig
Vorgeschlagene Pakete:
isag
Die folgenden NEUEN Pakete werden installiert:
sysstat
0 aktualisiert, 1 neu installiert, 0 zu entfernen und 0 nicht aktualisiert.
Es müssen 218kB an Archiven heruntergeladen werden.
Nach dieser Operation werden 901kB Plattenplatz zusätzlich benutzt.
Hole:1 http://ftp.de.debian.org lenny/main sysstat 8.1.2-2 [218kB]
Es wurden 218kB in 0s geholt (371kB/s)
...Udo
That's OK now,
I've used iostat command and it's output,
Linux 2.6.18-2-pve (vp) 04/26/10 _x86_64_
avg-cpu: %user %nice %system %iowait %steal %idle
0.31 0.00 0.93 1.11 0.00 97.65
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
hdf 0.02 3.50 0.00 198288 0
sda 23.63 809.89 1553.44 45830142 87905808
sda1 0.00 0.03 0.00 1490 56
sda2 23.62 809.87 1553.44 45828540 87905752
dm-0 0.01 0.01 0.10 376 5728
dm-1 38.73 213.30 297.15 12070178 16814896
dm-2 165.57 596.55 1256.19 33757514 71085160
So how can I determine I/O rate per vm now?
I've used iostat command and it's output,
Linux 2.6.18-2-pve (vp) 04/26/10 _x86_64_
avg-cpu: %user %nice %system %iowait %steal %idle
0.31 0.00 0.93 1.11 0.00 97.65
Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
hdf 0.02 3.50 0.00 198288 0
sda 23.63 809.89 1553.44 45830142 87905808
sda1 0.00 0.03 0.00 1490 56
sda2 23.62 809.87 1553.44 45828540 87905752
dm-0 0.01 0.01 0.10 376 5728
dm-1 38.73 213.30 297.15 12070178 16814896
dm-2 165.57 596.55 1256.19 33757514 71085160
So how can I determine I/O rate per vm now?
iostat will give you per block device statistics.
If you have multiple guest images on one block device, you won't see much.
In other words: you will see stats per guest if each "guest disk" is placed on a separate LVM volume.
If you have guest images in files (i.e. qcow2) placed on one mount point, you won't get data which is interesting to you - in this case, you can try "iotop".
If you have multiple guest images on one block device, you won't see much.
In other words: you will see stats per guest if each "guest disk" is placed on a separate LVM volume.
If you have guest images in files (i.e. qcow2) placed on one mount point, you won't get data which is interesting to you - in this case, you can try "iotop".
Hi,
if you use lvm-volumes for a VM, you can find the right dm-device with
Code:
dmsetup infoE.G. at the following output
Code:
Name: fc_1_vg-vm--104--disk--1
State: ACTIVE
Read Ahead: 256
Tables present: LIVE
Open count: 1
Event number: 0
Major, minor: 252, 12
Number of targets: 1
UUID: LVM-fosNqBYjAB8g493FXcnEavmlUJeA91Hc0SQEkFh9hfA8Ja0Xe8xcvNPQD66JB3wa
Code:
iostat -dm 5 dm-12Udo