Partition lost due to heavy load?

[segil]

Hello sysadmins!


I have got a problem with a harddisk and hopefully you can help me. We supply industrial computers to our customers with two harddisks, one 80 gb and one 250 gb. The 80 gb disk is used for Windows and software, the 250 gb is used for data storage. The computers store data coming from a production proces at the customer.
The ammount of data varies, depending on the ammount of data coming from the production proces. The files are 200-500 mb in size, with a total of 5-10. So each time data is stored to the drive, about 1-10 gb is written continuously. After that, the data is read for further analysis.
The 250 gb data drive has one big partion, NTFS formatted on a Windows 2000 Professional SP4 system with the latest updates.

Now the problem: occassionaly the partition on the data drive is just gone, after Windows gives an error message that the drive is corrupt or unreadable. At the same time Windows logs a number of NTFS errors in the system log.
At that point the data drive is inaccessible. After a reboot the drive shows up in the Windows disk management and can be reformatted. After that, the partition is back and can be used again. Of course, all data on the drive is gone.
We have ran into this problem several times, on different computers. We have tested the drive with Windows chkdsk, and also with the diagnostic tools from the harddisk suppliers. No problems are found and everthing looks okay. So we don't think the drive is malfunctioning. But for some reason the partition on the drive gets lost under heavy load.

Could it be that Windows 2000 professional isn't designed for handling extreme ammounts of data transfer? Maybe the IDE controller or it's driver cannot coop with this amount of data transfer? Maybe a bug in the Windows NTFS driver?

Below are some additional specs of the system.

System chipset: Intel 865G
250 gb harddisk: Hitachi Deskstar T7K250 IDE, connected to the built-in IDE controller of the chipset
Operating system: Windows 2000 Professional SP4
Drive info: 250 GB is NTFS formatted with "write caching" DISabled! (due to power failures)

All help is appreciated.
Thnx.

 

[stduc]

I've never seen anything like you describe. However here's a few things you may want to consider.

Is there any chance the drives could be over-heating?

Next time this happens, rather than just re-formatting the drive, let SpinRite have a go (go to grc.com). If the drive has suffered andy bad sectors then SpinRite will find them and may recover the data. A re-format will simply mark those sectors as bad while re-formatting.

If you need the data back - try something like getdataback (google for it).

Common causes of data loss on HDD's that work after a reformat are:-

1. Power loss while writing.
2. Over heating
3. Bad data (IDE/SATA) cables
4. EMI - Any potential source of electrical noise in the vicinity? - Is the computer on a UPS to isolate any mains noise?
5. Bad RAM - run memtest86 to verify RAM.

 

[segil]

Hi Stduc,

thanks for the reply. To anwser your questions:

One of the computers is coming over tomorrow for further analysis... I will test it thoroughly.

The computers are warm, because of the narrow space that they are placed into. Normally we look at the cpu temp, but I will try to monitor the HDD temp as well. The CPU temp varies between 50-70 degrees.
Electrical noise is also possible, because the cabinet with the computers is placed near heavy machines with big engines... in the past we have done some tests to see whether our ethernet connections suffer from this, but that isn't the case... still, it is possible electrical noise could interfere.

Thanks for your help and I will report when I have the test results.

 

[Grenage]

I would have imagine that the electrical noise would cause the computer to crash, rather than display the same partition problem you are seeing repeatedly.

Is it possible to use another hard drive brand and model or operating system in a machine for testing purposes?


Carlsberg don't run I.T departments, but if they did they'd probably be more fun.

 

[segil]

Hi Grenage,

I think we have solved the problem. Like many times it was something completely different than I thought of...

It had to do with the 48 bit LBA addressing support in Windows 2000 SP4. This should be turned on for the support of 137+ gb drives and that was not the case (I thought it was)... strangely sometimes Windows did recognize the full 250 gb and format it... but when trying to read and write files to the drive, problems arrise.

And after enabling the 48 bit support, it seems everything is okay now. No more corruption or partitions lost!

So simple, but so strange this support isn't enabled by default!

 

[stduc]

Odd that - as you say & well spotted!

 

[segil]

In case anybody runs into this problem.. here are two links about this subject...

http://support.microsoft.com/kb/305098/EN-US/

http://www.48bitlba.com/

 

[Grenage]

Thanks for coming posting back to let us know what the problem was, good to hear that you figured it out!


Carlsberg don't run I.T departments, but if they did they'd probably be more fun.