If anyone else is running the Adaptec 2805:  I believe there is a compatibility issue between the EARX (not EARS) drives.  When I purchased my last round of 2TB drives, I got home to find out they were EARX not EARS.  I plugged them in and started preclearing them only to find the drive speed was abysmal (started at 110MB/s and quickly wound down to less than 50MB/s).

 

After a couple of hours with them in the system, the array started getting unstable (read/write errors).  Once they were taken out, the array was stable.

 

I just ran across some people having issues with the new WD RE4 drives (enterprise) on the Adaptec controllers.  The workaround suggested by Adaptec was to change the PHY setting in the controller from 'Auto' to 1.5 Gb/s.  I made that change and have started a preclear -- the speeds are stable in the 105 MB/s range.

 

Will keep this thread updated as I go along and if I can get the system to be stable I will remove the caution.

 

Versions:

Unraid: 5b12

2805: 5.2-0[17899] (do need to update -- waiting on stable state)

EARX: WD20EARX-00P    51.005

 

 

UPDATE (7 Oct 2012):

 

I've been running this configuration for over a year now and I'm up to 5 drives on the Adaptec though I *HIGHLY* recommend against using this card.  As the betas have come along, the performance problem has been fixed but this card doesn't like idle WD Green drives (not sure if it's the intellipark issue or something else) so recoverable read errors happen quite frequently when the drives are woken by the card. 

 

Even though there are read errors, my data is still correct:  I do sha1 sums before uploading and I compare the sums to the files when I do a parity check.  The parity check passes and the sha1s pass.

 

I get a redball if the drive isn't awake when writing starts.  It's important to spin up all drives before stopping the array as just stopping the array tends to give me a redball.

 

The redball issue was especially severe when I had a Windows machine with SMB mounts to my unraid system:  something in Windows wants desperately to write small updates to the file system while mounted.  This was causing a redball once per week.

 

Every time I get a redball, I have to inspect the disk to make sure it's ok (no SMART issues).

 

I will be building a new array soon so that I can get rid of this card.

 

 

 

 

 

 

 

 

 

 

The preclear ran successfully but one of the EARS drives reported a few read errors.

 

Going to try to update the kernel on the card sometime this evening.

No good -- after about 20 hours one of the disks started showing some read errors.

Next step is to upgrade the kernel on the card

No good -- after about 20 hours one of the disks started showing some read errors.

Next step is to upgrade the kernel on the card

 

never rule out a bad batch of drives. it does happen.

 

Have you tried the same drives on a different controller?