Problem on debian with KVM / Disks

[aureld]

Hi all !
I'm new on this forum, but i'm using proxmoxVE since 6 or 7 months ago !
But now we have a little problem with one server :
It's a debian etch on KVM.
kernel on the vm' : 2.6.24-etchnhalf.1-686
kernel proxmox :2.6.24-2-pve (the last i think)

And we have problem with disk , for example this morning :

Apr 9 06:42:48 z kernel: hda: dma_timer_expiry: dma status == 0x21
Apr 9 06:44:14 z kernel: hda: dma_timer_expiry: dma status == 0x21
Apr 9 06:44:14 z kernel: hda: DMA timeout error
Apr 9 06:44:14 z kernel: hda: dma timeout error: status=0xd8 { Busy }
Apr 9 06:44:14 z kernel: ide: failed opcode was: unknown
Apr 9 06:44:14 z kernel: hda: DMA disabled
Apr 9 06:44:14 z kernel: ide0: reset: success
Apr 9 06:48:21 z kernel: hda: dma_timer_expiry: dma status == 0x21
Apr 9 06:49:05 z kernel: Clocksource tsc unstable (delta = 100606509 ns)
Apr 9 06:49:05 z kernel: Time: acpi_pm clocksource has been installed.
Apr 9 07:02:32 z kernel: hda: dma_timer_expiry: dma status == 0x21
Apr 9 07:02:32 z kernel: hda: DMA timeout error
Apr 9 07:02:32 z kernel: hda: dma timeout error: status=0xd8 { Busy }
Apr 9 07:02:32 z kernel: ide: failed opcode was: unknown
Apr 9 07:02:32 z kernel: hda: DMA disabled
Apr 9 07:02:32 z kernel: ide0: reset: success
Apr 9 07:03:27 z kernel: hda: dma_timer_expiry: dma status == 0x21
Apr 9 07:11:38 z kernel: hda: dma_timer_expiry: dma status == 0x21

We have this problem since a few days, and i already have tried to convert scsi disk inide disk (no we use ide disk), and it's the same problem...

Have you an idea of the problem ?

THanks a lot !

 

[tom]

1. What harddisk/raid do you use on the Proxmox VE host?
2. Pls post the results of 'pveperf' - but make sure that the host is not under load.
3. Pls post your VM config file (cat /etc/qemu-server/VMID.conf

 

[aureld]

1. What harddisk/raid do you use on the Proxmox VE host?
2. Pls post the results of 'pveperf' - but make sure that the host is not under load.
3. Pls post your VM config file (cat /etc/qemu-server/VMID.conf

Thanks for your answer

1. It's a single disk (no raid, etc...) on the host
THe hard disk :

root@xx# hdparm -i /dev/sda

/dev/sda:

Model=ST3750330AS , FwRev=SD15 , SerialNo= 5QK0E885
Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs RotSpdTol>.5% }
RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=4
BuffType=unknown, BuffSize=0kB, MaxMultSect=16, MultSect=?16?
CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=268435455
IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120}
PIO modes: pio0 pio1 pio2 pio3 pio4
DMA modes: mdma0 mdma1 mdma2
UDMA modes: udma0 udma1 udma2 udma3 udma4 udma5
AdvancedPM=no WriteCache=enabled
Drive conforms to: unknown: ATA/ATAPI-4 ATA/ATAPI-5 ATA/ATAPI-6 ATA/ATAPI-7

* signifies the current active mode


2. The pveperf :

root@xx:/etc/qemu-server# pveperf
CPU BOGOMIPS: 9336.41
REGEX/SECOND: 439875
HD SIZE: 9.61 GB (/dev/sda1)
BUFFERED READS: 70.62 MB/sec
AVERAGE SEEK TIME: 8.51 ms
FSYNCS/SECOND: 527.80
DNS EXT: 28.99 ms
DNS INT: 1.93 ms (xx.com)


3 .

ostype: l26
memory: 2500
onboot: 1
name: zz.xx.com
smp: 2
vlan0: rtl8139=B62:43:27:18:68
bootdisk: ide0
boot: dc
freeze: 0
cpuunits: 1000
acpi: 1
kvm: 1
ide0: vm-104-disk.qcow2

 

[tom]

pls check if you also got disk time outs on the host itself - maybe this is the beginning of hardware failure.

another question: is this a IO intensive KVM guest? maybe this sata disk is just no capable to handle the requests.

if you do not find any hardware problems you can convert the disk image file from qcow2 to raw format (with qemu-img), maybe this can help here - or better: upgrade hardware raid controller with BBU

 

[aureld]

pls check if you also got disk time outs on the host itself - maybe this is the beginning of hardware failure.

I don't have any disk timeout on the host. I will plan a disk change.

another question: is this a IO intensive KVM guest? maybe this sata disk is just no capable to handle the requests.

There is just a mail server on the guest (with zimbra). But i have some rsync script , and i'm going to stop them, and we will see...

if you do not find any hardware problems you can convert the disk image file from qcow2 to raw format (with qemu-img), maybe this can help here - or better: upgrade hardware raid controller with BBU

ok !


The big problem is that actually the guest is sometimes freezing : no more connectivity with the net, completely freez in kvm, and there is only one solution : reset the guest.

 

[tom]

just to mention:

I highly recommend to run Zimbra as a container, not as KVM.

As on openvz you do not have a virtualized disk you got MUCH better performance.

see http://pve.proxmox.com/wiki/Zimbra

 

[nick]

Sorry but Zimbra can not be run in OpenVZ emulation! I already try this! I receive error when zimbra start!

I try on 3 templates: Debian, Ubuntu and CentOS5 - all 3 give the same error!

On full KVM no errors!

I think these errors are regarding Java VM

My opinion...

 

[thefool808]

I run zimbra with openvz. I run the zcs-5.0.2_GA_1975.DEBIAN4.0.20080130234700.tgz package on a debian-4.0 template (I haven't updated it since I installed it, I'm a bad person >.<). I think I originally had an error where the vz was starting postfix (or something else taking port 25), but that was easliy resolved. It can be a RAM hog though (just like anything else running on a big java stack).

 

[nick]

I run zimbra with openvz. I run the zcs-5.0.2_GA_1975.DEBIAN4.0.20080130234700.tgz package on a debian-4.0 template (I haven't updated it since I installed it, I'm a bad person >.<). I think I originally had an error where the vz was starting postfix (or something else taking port 25), but that was easliy resolved. It can be a RAM hog though (just like anything else running on a big java stack).

I'm not sure if is the correct place to discuss about Zimbra but I will test also version 5.0.2. Personal, I make the tests with versions 5.0.13 and 5.0.14. Now, I use Zimbra on KVM with CentOS 5.2.

I will post the results and errors.

 

[tom]

I'm not sure if is the correct place to discuss about Zimbra but I will test also version 5.0.2. Personal, I make the tests with versions 5.0.13 and 5.0.14. Now, I use Zimbra on KVM with CentOS 5.2.

I will post the results and errors.

Zimbra runs VERY well on OpenVZ, just follow our wiki page and the Zimbra docu. If you encounter any issue just post it in a new thread.

 

[nick]

I read the Wiki but nothing special! I will test on Ubuntu to see what happen! It's strange why it doesn't work on my previous tests!

Zimbra can be used with other templates or only with Ubuntu?

 

[tom]

I read the Wiki but nothing special! I will test on Ubuntu to see what happen! It's strange why it doesn't work on my previous tests!

Zimbra can be used with other templates or only with Ubuntu?

You can choose also others but Ubuntu is a supported platform for Zimbra (also for the commercial editions).

 

[nick]

I will post the results and errors.

here are the results: http://www.proxmox.com/forum/showthread.php?t=1307