PERC6 drops WD

[joeWI]

I have 2 DELL PowerEdge 2900 servers with PERC6i (RAID) controllers. Both servers have 2 HDDs in a RAID-1 mirror for the O/S.

Server A uses 6 WD 7200RPM RE3 1TB drives in a RAID-5 array for data and runs fine.

Server B uses 6 WD 10K RPM Velociraptor 300GB drives in a RAID-10 array for data and a few virtual machine vhd's and has problems. Anywhere from 1 to 10 days, it will drop a drive from the RAID array. Almost always from the RAID-10 "data" array.

I used the WD TLER utility to check the drives. Both sets of drives have TLER turned on and set to 7 seconds read and 7 seconds write.

DELL has replaced the PERC6i with a newer version (new firmware, bios, etc.) and it still is dropping drives.

On several sites, I have read the the Velociraptors push data so fast that it overwhelms the RAID controller and that can result in a dropped drive.

Does anyone know if that is true?
Should I change the TLER value to something shorter like 5 or 6 seconds?
Should I replace the Velociraptors with more RE3's?

Thanks in advance for any information that you can provide.
Joe

[venger]

I also read something, I think on newegg reviews, that the non RE drives do some sort of periodic integrity check or whatever and that causes a drop. I think WD's response regarding a fix was to use the RE version of the drives...

[joeWI]

Thanks for the reply.

I too have been reading a lot over the last week and have come to the conclusion that the Velociraptor drives only work reliably with a select few RAID controllers.

Western Digital appears to be unwilling or unable to fix the drives for broader compatiblity.

Unfortunate for me, the DELL PERC6i is not one of them so I'll have to replace all the hard drives.

Due to the number of posts, I'm not the only one that has fallen into this GOTCHA situation and Western Digital's silence is sad.

[joeWI]

Update....

Western Digital has offered to replace the hard drives with more of the same.

Dell suspects that the RAID virtual disk is corrupt and I should recreate it. So I backed up all data, deleted the RAID-10 virtual disk, created a new RAID-10 virtual disk, executed a full initialization, created a basic disk primary partition, formatted the partition, restored all data.

It's been 10 days so far and no dropped drives. YEA!

I have been looking into how the virtual disk definition could of become corrupted and have the following observations and theory.

By moving the virtual disk between PE2900's, it somehow became corrupted. (Both servers were always powered down.) Both PER6i controllers have battery backup with write-back enabled. (That's the default.) In reading the PERC6i manual, it states that the battery backup will keep any unsaved data and when the virtual disk in inserted at a later time and the foreign configuration is accepted, it will write the data. BINGO. I have corruption.

To prevent this from happening, I have chosen write-thru which writes to the physical disks before responding write completed.

Long story short....yea I know too late...I did this to myself. But on a graceful powerdown, I would have thought that the RAID controller would have been sent a signal that the server is powering down and flush everything to the physical drives.