NT4 vm freezes on heavy disk load

I had the job to covert 3 old physical NT4 Servers into Proxmox VM’s cause theay are running a native production software only available under nt4.

Two of them are running stable now and I am very happy with it.

The third one is even running most of time, but has an issue on heavy disk I/O. After starting a log checker, a statistic tool or something else (like chkdsk) which produces heavy load on the disk, the vm freezes.

Booting, restarting and copying of a few small files results in no problems, copying of big amounts of files or big files results in freezing.

The only status messages I can see is ‘Status: io-error’ on the Proxmox web interface summary screen.

The both well running vm’s are having only one disk each of 4 GB size (running atapi.sys driver).
The third struggling vm has 3 three disks also running via atapi.sys. The first (bootdisk) has a size of 4GB, the second has a size of 26GB and the third one has a size of 5 GB.

The proxmox 2.2 configuration consits of a two node cluster with drbd. There are several vm's running stable (differents linuxes als well as w2k3 and w2k - and newly three NT4)

I am using:


here is the config of the vm:


I know, NT4 is very old and there isn't much experience in the world anymore, but I hope that some of you guys are having a hint for me where I can find something to get it running.

spirit said:

maybe try to add "cache=none" or "cache=directsync" to your drives.

Click to expand...

Hi spirit,

thank you for your reply.

I've tested both cache entries on all drives but with no better results.

udo said:

@ewuewu
Other question, if the NT with one disk is stable - can you put all data togetter on one huge disk? (in old days you can do assign other drives (d->c; e->c))??
Perhpas the software will work also with this config?

Udo

Click to expand...

Hello Udo,

even writetrough shows the same behaviour.
I'am not really sure that the 'assign' trick will work in this case, because even high load the boot disk 'c:\' lets the system freeze. But I will give it a try. Hope this works cause the programmer of the software wrote a lot of hardcoded pathes in into his code.

Thank you so far

... maybe I have to lock anywhere else. Can somebody explain what 'Status: io-error' in the web interface means in detail? I can't find any according messages in syslog of the vm host or in the eventviewer of the NT4 guest.

Is there another place where I can find detailed log messages?

Maybe I had made mistake during the p2v process. But I used the same procedure on the other 2 well running vm’s.

The old physical machine had an old adaptec scsi raid controller. (same on the other both old machines)

First in the process I’ve installed the atapi.sys driver and reinstalled sp6 afterwards. (still on the physical machine)

Then I’ve made a p2v with vmware converter 4

In the next step I converted the split vmdk files of each disk into flat files by using vmware-vdiskmanager.

Afterwards I created a vm on the proxmox host with 3 ide disks of the right size (located in the LVM space of my mirrored drbd drive).

After this I did a dd of each flat file into the according LVM-disks.

Last step was to boot the vm.

Up to this step everything was working fine and the vm was running for about one day.
In the following night the system was freezing. The cause was the high load on the statistic tool.

Afterwards I was searching for the problem and I could reproduce this by running chkdsk or copying a large bunch of files or simply executing a backup into a network drive.
The freezing occurs on operations on each disk. But only on high load.
If I stoped the statistic tool, the vm was running for several days. If I do a copy file by file from or to the vm the vm works as expected.

spirit said:

maybe can you try with scsi disk ?

performances are better than ide, and I think that the lsi controller should work with nt4

Click to expand...

That could be a good idea. Do you have any I idea which driver I should use in NT4 vm? There are only a few drivers availabe.