My hard drives are well out of warranty. I’ve been thinking about that for a long time now with a bit of dread. I’ve had plenty of hard drives last longer than five years but my file server has been approaching that mark and I’ve felt less than comfortable in seeing how good my RAID5 really is. I do have backups elsewhere, this is true. But they aren’t always fresh and I know I’d regret loosing something or another. A stable file system is just something I need to have and there’s no excuses not to have one.
My file server is a beast. It’s sitting in an old cube case I had hacked to bits for a water cooling project many years ago and is the size of a short end table. If the size of the thing isn’t a theft deterrent than the weight of it surely is. I’d be tempted to fill part of the case with concrete to make it even less desirable but I figure that I’d be the one having to eventually lug it to a new apartment.
Right now it’s an old Asus P4C800 and P4 3.2C with some good PC4000 RAM, and Audigy 2, a FX5200, and a Hauppage dual-tuner capture card. Basically all of my left overs and remnants from years past. This machine is more than a file server actually since I use it as my main workstation as well. I have a better machine but that one’s solely for overclocking and gaming. If I want to watch movies, check email, or if my other machines are down for the count, this is my baby. My last resort. The forgotten.
It hasn’t been cleaned in two years and since it sits on the floor its a real mess. I’ll have pictures of this later.
It sports EIGHT 250G drives. Six Western Digitals at nearly 5 years of age. Two Maxtors less than two years old, purchased to replace two of the WD’s that fell out of the RAID5 array and wouldn’t go back. Strangely enough, the two ‘dead’ WD’s started working again after I ran drive fitness on them. I reconnected them to the machine as a simple JBOD backup set which I stored downloaded video on, nothing of importance.
One of those two started making horrible grinding noises and died completely two weeks ago, spurring me on to rethink this entire situation.
Those two drives were JBOD as my Q: drive, worth 500GB. I have two WD’s running off of the onboard Promise controller in a 500GB RAID0. This runs the C: (boot and swap), E: (OS) and O: (Video & Downloads) drives. Nothing of importance stays on the RAID0 array and I keep around simply for speed. I would have originally added this to the X: drive array but it was full. Speaking of the devil, X: consisted of four 250GB drives in RAID5 mode giving me around 750GB of redundant storage. Not bad in 2003 or 2008. This was on a 3Ware 7500-4 RAID controller. Really good for the day.
All of these drives, except the two in RAID0, are PATA drives. This complicated the upgrade path since I really wanted to go SATA, and needed to do this so I could get the larger drives. Larger drives because having EIGHT drives at once is noisy and a greater risk of failure than having fewer drives. Also this machine is on 24×7 and anything I can do for power savings is worth it.
Cost was no object five years ago but it was a huge object this time. I couldn’t afford to get the best drives and RAID controller this time. A fact which might bite me in the ass sooner than later. What enabled me to finally do this upgrade was a beautiful Amazon gift certificate I got for the holidays from work. There are a couple PC vendors you can buy through on Amazon and the prices are actually competitive. But the selection is smaller than I would have liked.
I wanted Seagates but was forced to go with Western Digital Caviar WD10EACS drives. These are desktop drives with only a three year warranty and 5400RPMs. But they are CHEAP and have 1TB of storage each. I only needed two but RAID5 requires at least three drives. RAID5 has the benefit of really decent read speeds and also only costs one drives worth of space for the redundancy, in this case 33% of the space. RAID1 mirroring costs 50% and doesn’t have any performance benefits.
I needed a SATA RAID controller and here 3Ware and other top end brands hit the $300 mark quickly. I went with a much more affordable Highpoint 1740 controller. It has all the features I needed, RAID5, PCI (not PCIe), etc. It’s not a true hardware RAID so the CPU is going to get abused a bit more. On the other hand it does support online capacity expansion which will be great if I decide to add a forth 1TB drive to my array. I regret not being able to get a PCIe 4x version as the extra bandwidth there would have been nice but I have no plans to upgrade the rest of this PC anytime soon.
Installation is ongoing. It took over 24 hours to initialize the 2TB RAID5 array and another 6-8 hours to copy over the 600GB of data from the old RAID5 array to the new one. I’ve unplugged the old four drives and controller now and everything seems to be working fine.
But I’m worried. You see, WD10EACS drives aren’t meant to be run in RAID arrays. Apparently if the drives encounter a data error they will enter a ’scrub’ cycle designed to help recover any lost data. This forces the drive to time out and causes the RAID controller to drop the drive. This is according to Western Digital. They recommend that you spend $20 more, get their RAID version drive which has firmware set to have a smaller ’scrub’ cycle. If there’s any other difference to the drives I have no idea. Unfortunately the option to get the RAID drives wasn’t available to me due to limited selection and funds. There are reports of other people having no problems with these drives in RAID. So far I have not had any. I hope that the worst case scenario for me is that I loose one drive and I may need to rebuild the array, but I should have no major faults. If two drives fall out of the array at once though I could be screwed.
There might be a way to hack the firmware of course. Maybe WD will come to their senses and make this an ‘option’. Either way, I knew I should have gone with another brand. Beggers can’t be choosers though.
What will happen to the old drives? I got a few USB enclosures and a network share box for them. One will find its way to my MP3 box at work. None of that will happen until I’m sure that my data survived the transfer and that this new array isn’t going to kerplode anytime soon. I wonder how long I should wait?