DOA Hard Drive

Friend at a medium size corp, stated they run into about 1 bad drive out of every 100 they purchase.

Our cinema servers at film festivals used 1-2GB HD/4-drive array RAID-5 w/automatic raid rebuild even during high data playback. Had some drives fail, though never resulted in loss of presentation. Was very reliable.

I built a RAID-5 for 4K editing connected by Thunderbolt to my iMac i9 at work. Also, very reliable.

theoldwizard1 wrote:

rlw999 wrote:
I wouldn't run RAID-5 with such a large array, at least not unless you have good backups or don't care if you lose the data.

RAID is NEVER a substitute for backup !

Admittedly, my RAID array only gets backuped up once a week.

I agree, that RAID is not backup, but I also know that most home users don't have a reasonable way to back up 20TB+ of data. Thanks to a 1TB bandwidth cap from my ISP, it would take me nearly 2 years to back up or restore that much data unless I'm willing to pay a huge bandwidth fee.

I have less than a TB of data that I consider to be important, and I back that data up with a cloud backup service. The rest is mostly movies and I have the original disks for them, so I'd just re-copy them if I had to. I have most of the data backed up on a external hard drives, but they aren't stored off-site (but are at last stored on the opposite side of the house from the fileserver), so aren't great back ups.

Gdetrailer wrote:

rlw999 wrote:

philh wrote:
First time I've ever had a hard drive DOA. Was upgrading drives in my home server, as the drives were a bit old and wanted to increase space. Five years ago, said I'll never run out of space, LOL. Four 8TB raid 5 should be good for awhile!

Large RAID-5 arrays take a long time (a day, or multiple days) to resilver after a disk loss, and if you lose another disk during the rebuild, you'll lose the entire array. Rebuilding itself stresses the disks making a multi-disk failure not uncommon.

Have worked with RAID systems on servers since the 1990s and what you posted isn't really true on good quality RAID systems.

Good quality dedicated hardware based RAID controllers have their own processor and on board buffer memory with a backup battery and handles rebuilds in the background and you never see any data loss with RAID5 unless you lose more than 1 drive at a time, perhaps might see a slight reduction in array speed while array has be degraded.

Didn't you just say the same thing as me but you took the optimistic view?

I said that with RAID-5 if you lose another drive during the rebuild, you'll lose the entire array, you said that if you don't lose another drive during the rebuild, you won't lose any data.

Whether you have hardware RAID or not, resilvering 24TB of data is going to take days, maybe over a week. And during that time the disks are going to be 100% busy so if any of the other ones were near failure, that's when they are going to fail.

Back when "large" drives were 150GB SAS drives that could sustain 200MB/sec (with a deep 254 entry command queue), rebuilds happened relatively quickly and a hot spare could save the day. But throughput hasn't kept up with capacity increases, now you can get 8TB drives (50X larger) and you're luck if you can sustain 150MB/second. Just reading 8TB at 150MB/second takes 14 hours, but when you're rebuilding a RAID array you're reading and writing at the same time, so throughput drops.

Best to avoid software based RAID, just not robust or fast once you see the difference between a dedicated Hardware based RAID.

I run ZFS RAID-Z2 (which is close to software RAID-6) and one nice advantage is that when it has to resilver the array, it knows which blocks have data so it only needs to rebuild parity for that data, so if I'm only using 1T out of 8TB, that's all it needs to rebuild.

JaxDad wrote:
I’ve dodged the bullet so far, but........ it’s going to happen, soon.

The big issue, and one of if not my biggest P!&% off, is the external drive DVR on my satellite TV service.

Some clever bunch of corporate bean-counters decided that if folks could access the content they’d have a data sharing problem.

Their solution? When you plug a new drive in to the sat. box it formats it in Linux and installs an auto-run program that looks at the S/N of the box, if it doesn’t match it won’t play. It also uses some weird, even for Linux, encryption that means you can’t even duplicate the drive.

So with MANY hours of content we can’t protect it, if the drive dies, it’s gone. It’s already noisy...... tick, tick, tick.

JAX, you need to thank the MPAA (Motion Picture Association), screen writers guild, actors guild for the encrypted DVR drives on sat TV boxes. The decision was not made by the Sat providers but by the movie and TV industry.

If the content is not encrypted (protected) then the Sat companies cannot prove that they were not involved with pirating the content after it is recorded on the DVR drive.

On some of the older Dish receivers, there is a USB port that will allow you to connect an external USB HD that allows you to copy DVR content to the drive. You do have to pay a one time fee for Dish to activate that feature for that unit.

I have not tried that but I do believe if you get a replacement unit (same model) you should be able to plug the USB drive in and it should be able to play the content on that drive. You may need to pay another one time fee for activating the replacement receiver USB port.

External USB drives will be reformatted and the data placed there from the DVR will be encrypted so you can't view or copy content digitally via a computer. But in theory you most likely should be able to view the content of the USB drive on a replacement receiver.

Now, if you were just try to transplant or copy/image the internal receiver DVR drive, yeah, those are "locked" to that receiver SN so transplanting in that case isn't going to work.

theoldwizard1 wrote:

rlw999 wrote:
I wouldn't run RAID-5 with such a large array, at least not unless you have good backups or don't care if you lose the data.

RAID is NEVER a substitute for backup !

Admittedly, my RAID array only gets backuped up once a week.

Correct!

While RAID levels above RAID 0 provide redundancy of the drives themselves, it however, does not provide redundancy for the RAID controller hardware nor the PC hardware.

A failure of the RAID controller can deal a fatal blow to the data on the drives especially while the drives are being written to.

Additional backup methods are essential for any data you deem important or critical.

Good practice is using several types of backups for good measure, for personal use, a simple external USB hard drive will work fine but should consider one step further and use at least two different external USB drives. Alternate between the drives every other backup.

I have had external USB hard drives fail which is why two different USB drives are important..

rlw999 wrote:
I wouldn't run RAID-5 with such a large array, at least not unless you have good backups or don't care if you lose the data.

RAID is NEVER a substitute for backup !

Admittedly, my RAID array only gets backuped up once a week.

I’ve dodged the bullet so far, but........ it’s going to happen, soon.

The big issue, and one of if not my biggest P!&% off, is the external drive DVR on my satellite TV service.

Some clever bunch of corporate bean-counters decided that if folks could access the content they’d have a data sharing problem.

Their solution? When you plug a new drive in to the sat. box it formats it in Linux and installs an auto-run program that looks at the S/N of the box, if it doesn’t match it won’t play. It also uses some weird, even for Linux, encryption that means you can’t even duplicate the drive.

So with MANY hours of content we can’t protect it, if the drive dies, it’s gone. It’s already noisy...... tick, tick, tick.

rlw999 wrote:

philh wrote:
First time I've ever had a hard drive DOA. Was upgrading drives in my home server, as the drives were a bit old and wanted to increase space. Five years ago, said I'll never run out of space, LOL. Four 8TB raid 5 should be good for awhile!

I wouldn't run RAID-5 with such a large array, at least not unless you have good backups or don't care if you lose the data.

Large RAID-5 arrays take a long time (a day, or multiple days) to resilver after a disk loss, and if you lose another disk during the rebuild, you'll lose the entire array. Rebuilding itself stresses the disks making a multi-disk failure not uncommon.

I run a 5 disk RAID-6 array at home, which means I can lose up to 2 disks without data loss, but when I upgrade it, large disks are so cheap that I'll probably just go with mirroring (which means much faster rebuilds)

Have worked with RAID systems on servers since the 1990s and what you posted isn't really true on good quality RAID systems.

Good quality dedicated hardware based RAID controllers have their own processor and on board buffer memory with a backup battery and handles rebuilds in the background and you never see any data loss with RAID5 unless you lose more than 1 drive at a time, perhaps might see a slight reduction in array speed while array has be degraded.

Built on MB software based RAID controllers are however an whole different issue, this type depends on using your PCs processor and memory plus software to emulate the RAID controller. This can cause speed and possible data loss during a rebuild and you will have a very noticeable array access speed reduction during rebuild.

Good quality hardware based RAID systems are expensive but will out perform software RAID systems.

With a dedicated hot spare plus a dedicated hardware RAID controller I kept a production server alive and running for 15 yrs before the server MB suffered a hardware failure.

During that time the server was up 24/7/365 and only shut down for a few power outages that outlasted the UPS and a couple of rack moves.. Served up Ghost images well over 20gig in size and imaged north of 25,000 desktop and 10,000 servers during it's life.. Wore out 3 SAS server rated drives (OS boot array) and 8 1.5TB non server rated SATA drives (data storage) during its life.

Never had to shut down the server for a drive change, the hot spare would automatically be inserted into the array when a drive failed and rebuild happened automatically. The only way I knew that happened was to either visually check the drive lights status or check the drive array controller via remote desktop.

For personal use, anything over RAID 1 (mirrored set) is a bit over the top unless you want to take advantage of expanding your drive space in RAID levels above RAID 1.

Never use RAID 0, no redundancy and if one drive fails all of your data will be lost across all drives in that array..

Best to avoid software based RAID, just not robust or fast once you see the difference between a dedicated Hardware based RAID.

I have lost two drives. One catastrophic (Thankfully nothing on it was all that important. Couple of downloaded videos and some managment software that was free)
And one ... Well I had a hit so I did a FULL backup... that was the last successful operation. Bullet dodged.

Now days it's all backed up across multiple devices.. A full loss is all but impossible.