Hi mates

 

I just found one strange behaviour. Since a few months ago I have an HP Microserver NL 36, AMD Athlon II NEO N36L dual core 1,3Ghz, 8GB DRR3 RAM no ECC, with 3 WD20EARX and unraid 5.0 beta 14 basic. Last monday I bought 2 WD30EFRX. I stopped the array and replaced the parity disk with a WD30EFRX, and it took 25 hours to rebuild the parity, IO wait meanwhile is always 60 or 70 IO wait. That looks strange to me, but I dont worried to much thousands of read's from both data disk and thousands of writes in the parity disk.

 

On Thursday I began to preclear the second WD30EFRX, a single full cycle took 35 hours, again an enormous IO wait, also erratical throughput, sometimes 145 mbs, sometimes 110 mbs, but most of time just 45 mb/s. Again my thought was that something really strange its happening. But other people are reporting that in the preclear results thread that 3TB disks, from others vendors also have 45 / 50 mb/s throughput.

 

This afternoon. I've decided to make a few tests, so I take out of the array both WD30EFRX, and rebuild the parity in the original WD20EARX. What's my surprise when the throughput it's constant 110 mb/s  or 112 mb/s all time, an insignificant IO wait (less than 2%) . And the full parity rebuild less than 5 hours.

 

Tomorrow I will start a new array with only both WD30EFRX in the internal controller, and after that I will try the same procedure but both disks in an ASUS U3S6 Dual USB3.0 + DUAL SATA 3 PCIe x4 controller.

 

I suspect that the Microserver's BIOS, can adress 3Tb hard drives, but it has problems with the mandatory LBA 64 that allows full support of 2.2 hard drives or don´t like WD30EFRX variable smart rotational speed (5900 to 7200 RPM). Also I will talk WD support if I can force a fixed rotational speed.

 

Has somebody found something similar with Microserver  + 3 TB disks? Could be a bug of the Unraid 5.0 B14?

 

I'll keep you informed of my progress.

 

Cheers

I've noticed a similar drop in speed during parts of the preclear process. Same drive but not using it in a Microserver.

 

I've also got a Hitachi 5K3000 that I pre cleared a while ago, took around 36 hours, where as the WD Red took around 34 hours. The speed also dropped while pre clearing the Hitachi drive.

 

Is it not just a case of drives not being able to maintain the same read/write speed across the entire surface of the disk?

 

Can't say I've really tested it but I thought it was to be expected really. Be interested to hear what others have to say on this topic.

 

Sent from my Samsung Galaxy S2 using Tapatalk 2