Network copied files are corrupted on Promise RAID5 array

[idon.wong]

Hi all,

So I just put together the following system and am encountering some
weird stuff with respect to my RAID 5 array. I decided to use a
dedicated Promise controller so I disabled the onboard SIL3114 Raid
controller.

AMD Athlon 64 X2 4200+
Asus A8N-SLI Deluxe
OCZ PC3200 DDR400 Premier Series 1 Gb Dual Channel
Gigabyte GeForce 6600LE 128MB PCI-E
Antec SLK3000B Mid Tower
Antec TruePower II 480W
1x Seagate ATA 200Gb

3x Western Digital SATA 320GB SATA drives
Promise FastTrak S150 SX4 with 64Mb DIMM
Microsoft Windows XP Pro

I use the 200Gb drive as my main O/S drive in a non-raid configuration.

Basically, I've done the following:
1. Create RAID5 array with the 3x320Gb
2. Initialized for RAID 5 the array
3. Formated a 200Gb NTFS partition

Here's where the problems start:
4. Copy files (about 15MB in size) from a network share to the raid
partition.
5. MD5 sums differ for the source and destination files.
6. Furthermore, repeated MD5s on the same files on the RAID give
differing results.
7. Copy the files from the RAID to a local non-raid hard drive
8. MD5 sums on the local drive files differ from the raid files.
9. Copying the files AGAIN from the RAID to the local drive yields
different MD5 sums again (differ from 8.).
10. Copy files from local to raid drive.
11. MD5 sums are identical, even after repeated copies.

These results seem to indicate that network copied files cause real
problems on the RAID5 array, that persist even after the files have
been written. 6. and 9. indicate even read operations seem to change
the data.

I have the latest bios, drivers and Promise Array Management for the
SX4 controller. I've tested the memory on the controller with
Promise's test utility.

I've contacted Promise support and am waiting for their response. Has
anyone ever seen anything like this before?

Thanks,
Don

 

[Paul]

Hi all,

So I just put together the following system and am encountering some
weird stuff with respect to my RAID 5 array. I decided to use a
dedicated Promise controller so I disabled the onboard SIL3114 Raid
controller.

AMD Athlon 64 X2 4200+
Asus A8N-SLI Deluxe
OCZ PC3200 DDR400 Premier Series 1 Gb Dual Channel
Gigabyte GeForce 6600LE 128MB PCI-E
Antec SLK3000B Mid Tower
Antec TruePower II 480W
1x Seagate ATA 200Gb

3x Western Digital SATA 320GB SATA drives
Promise FastTrak S150 SX4 with 64Mb DIMM
Microsoft Windows XP Pro

I use the 200Gb drive as my main O/S drive in a non-raid configuration.

Basically, I've done the following:
1. Create RAID5 array with the 3x320Gb
2. Initialized for RAID 5 the array
3. Formated a 200Gb NTFS partition

Here's where the problems start:
4. Copy files (about 15MB in size) from a network share to the raid
partition.
5. MD5 sums differ for the source and destination files.
6. Furthermore, repeated MD5s on the same files on the RAID give
differing results.
7. Copy the files from the RAID to a local non-raid hard drive
8. MD5 sums on the local drive files differ from the raid files.
9. Copying the files AGAIN from the RAID to the local drive yields
different MD5 sums again (differ from 8.).
10. Copy files from local to raid drive.
11. MD5 sums are identical, even after repeated copies.

These results seem to indicate that network copied files cause real
problems on the RAID5 array, that persist even after the files have
been written. 6. and 9. indicate even read operations seem to change
the data.

I have the latest bios, drivers and Promise Array Management for the
SX4 controller. I've tested the memory on the controller with
Promise's test utility.

I've contacted Promise support and am waiting for their response. Has
anyone ever seen anything like this before?

Thanks,
Don

I get dizzy, every time I try to follow exactly what operation
is being corrupted in this thread. Anyway, have a look and
see if any of this is familiar...

http://forums.nvidia.com/lofiversion/index.php?t8171.html

If you move the drives to a different motherboard, perhaps you
can use one of the file comparison applications those guys
are using, and find out which byte(s) are corrupted. Moving
RAID drives around will be tougher than moving single
disks, so it might not be that easy to test RAID volumes.

Paul

 

[Roger Hamlett]

Hi all,

So I just put together the following system and am encountering some
weird stuff with respect to my RAID 5 array. I decided to use a
dedicated Promise controller so I disabled the onboard SIL3114 Raid
controller.

AMD Athlon 64 X2 4200+
Asus A8N-SLI Deluxe
OCZ PC3200 DDR400 Premier Series 1 Gb Dual Channel
Gigabyte GeForce 6600LE 128MB PCI-E
Antec SLK3000B Mid Tower
Antec TruePower II 480W
1x Seagate ATA 200Gb

3x Western Digital SATA 320GB SATA drives
Promise FastTrak S150 SX4 with 64Mb DIMM
Microsoft Windows XP Pro

I use the 200Gb drive as my main O/S drive in a non-raid configuration.

Basically, I've done the following:
1. Create RAID5 array with the 3x320Gb
2. Initialized for RAID 5 the array
3. Formated a 200Gb NTFS partition

Here's where the problems start:
4. Copy files (about 15MB in size) from a network share to the raid
partition.
5. MD5 sums differ for the source and destination files.
6. Furthermore, repeated MD5s on the same files on the RAID give
differing results.
7. Copy the files from the RAID to a local non-raid hard drive
8. MD5 sums on the local drive files differ from the raid files.
9. Copying the files AGAIN from the RAID to the local drive yields
different MD5 sums again (differ from 8.).
10. Copy files from local to raid drive.
11. MD5 sums are identical, even after repeated copies.

These results seem to indicate that network copied files cause real
problems on the RAID5 array, that persist even after the files have
been written. 6. and 9. indicate even read operations seem to change
the data.

I have the latest bios, drivers and Promise Array Management for the
SX4 controller. I've tested the memory on the controller with
Promise's test utility.

I've contacted Promise support and am waiting for their response. Has
anyone ever seen anything like this before?

Thanks,
Don

What slot is the Promise in?. In the past, on an older board, I have seen
similar problems, when the controller is sharing an IRQ with the network
card, and the OS is running without extended interrupts enabled.

Best Wishes

 

[idon.wong]

The controller is in the last PCI slot. I'll check the IRQ assignments
tonight.

That's an interesting (and long) thread in the nvidia forum about
Nvidia's storage drivers causing data corruption. Not sure if it's
relevant but definitely something to try out. I think i remember
installing those. Though I've had no problems with the non-raid drive
yet.

Cheers,
Don

 

[CoCo]

(e-mail address removed) skrev:

The controller is in the last PCI slot. I'll check the IRQ assignments
tonight.

That's an interesting (and long) thread in the nvidia forum about
Nvidia's storage drivers causing data corruption. Not sure if it's
relevant but definitely something to try out. I think i remember
installing those. Though I've had no problems with the non-raid drive
yet.

Cheers,
Don

hi

i also had a simular thing, it was the nic!! driver setting namned "offload largesend" must
be off! (gigabit nic from d-link)

coco

 

[idon.wong]

Update after a few quick tests this morning:

1. Copying 10 x files (15Mb) from network drive to raid array works
ok.
2. Copying 40+ x files from network to raid doesn't work. Random files
are corrupted.
3. Copying 40+ x files from network to normal (local) drive works ok.
4. Copying 10 x files from local to raid works ok.
5. Copying 40+ x files from local to raid causes file corruption.
6. Using either network port on the A8N gives the same result.

It seems the problem is not limited to network copies as i first
thought. Looks like its the amount of data that's being copied to the
raid array that's the problem.

I'm gonna reinstall XP and NOT install the nForce ATA drivers to see
what happens next.

Cheers,
Don