Choosing File Storage/Backup Solution

[n00zler]

Sorry I should have clarified

WHS is Server2K3 with an application called WHS.....it fully supports GPT, as you can add a 20tb raid array to it if your heart desires. DE deletes the partition and creates a NTFS partition. It does not check to see what type of partition table you are using nor does it care.

So, I take it, WHS doesn't treat the storage pool like a single logical volume, but rather as multiple volumes working together. So as long as you use 2TB or smaller drives, you're fine. Or, you could use something other than WHS ot set up a RAID array or another logical drive and have WHS handle it as part of the storage pool.

"The WHS storage manager (QSM) formats disks using MBR, not GPT. This means that a single physical (or logical, in your case) disk can be a maximum of 2 TB in size." (http://forum.wegotserved.co.uk/index.php?showtopic=4502)

That's what got me a bit concerned, but I guess I was just misunderstanding what they were saying.

So, should I ever want to add a RAID array or a larger than 2TB drive, I just need to format the drive/array without using the storage manager. Correct?

 

[Ockie]

So, I take it, WHS doesn't treat the storage pool like a single logical volume, but rather as multiple volumes working together. So as long as you use 2TB or smaller drives, you're fine. Or, you could use something other than WHS ot set up a RAID array or another logical drive and have WHS handle it as part of the storage pool.

"The WHS storage manager (QSM) formats disks using MBR, not GPT. This means that a single physical (or logical, in your case) disk can be a maximum of 2 TB in size." (http://forum.wegotserved.co.uk/index.php?showtopic=4502)

That's what got me a bit concerned, but I guess I was just misunderstanding what they were saying.

So, should I ever want to add a RAID array or a larger than 2TB drive, I just need to format the drive/array without using the storage manager. Correct?

There are many ways around the 2tb limitation per drive size. Also, WHS is working on a new patch/release that will address this issue.

 

[brons2]

brons2: you worry about multiple drive failure within a limited time (24 hours), which is very very uncommon. More likely all drives fail due to power problems or letting your case fall sideways; causing a head crash on multiple disks at once. RAID6 does not protect you against that.

RAID6 also does not protect you against lost buffers. A single crash could cause large buffers to be lost, carrying metadata for your NTFS filesystem. NTFS works great on HDD buffers, but not on write-back buffers like RAID5/6. If it fails, it can bring with it large portions of the filesystem becomming inaccessible, even those who you have not read or written to in a long time. NTFS also does not have any means of detecting subtle meta-data corruption. NTFS is old.. but at least its not FAT.

So you can have multiple redundancy but still a corrupt filesystem. Just know RAID can never replace a backup; those two act supplementary. I'd rather have two RAID0 arrays keeping a copy of eachother, than one big RAID6 array. That said, the latter situation is much simpler to administer than a nightly backup system.

Enlightenment, I know this is probably the wrong thread, but I was talking about my work environment. There's no reason any of our hard drives would ever be dropped, as they're securely mounted in a rack. And there's no reason the power would go out, we have a 200Kva UPS and a generator, and every server has a minimum of 2 power supplies. I know, weird things happen sometimes, but the state of write buffers is not one of the things that keep me up at night.

That being said, yes, all our RAID cards that run SCSI or SAS direct-attached arrays have battery backed write cache. (Thankfully we are moving away from this with our virtualization and SAN projects so I won't have to have this discussion anymore).

But I have seen a 2nd drive fail in RAID 5 during the rebuild of a previously failed drive. At that point, the array is hosed, and I'm looking at my last backups. All the battery backed cache in the world won't save you in this case.

 

[Ockie]

Enlightenment, I know this is probably the wrong thread, but I was talking about my work environment. There's no reason any of our hard drives would ever be dropped, as they're securely mounted in a rack. And there's no reason the power would go out, we have a 200Kva UPS and a generator, and every server has a minimum of 2 power supplies. I know, weird things happen sometimes, but the state of write buffers is not one of the things that keep me up at night.

That being said, yes, all our RAID cards that run SCSI or SAS direct-attached arrays have battery backed write cache. (Thankfully we are moving away from this with our virtualization and SAN projects so I won't have to have this discussion anymore).

But I have seen a 2nd drive fail in RAID 5 during the rebuild of a previously failed drive. At that point, the array is hosed, and I'm looking at my last backups. All the battery backed cache in the world won't save you in this case.

Don't rely too much on your UPS and your Generator, they can fail. One of our competitors had a UPS system surge and blow up, literally, causing the suppression system to be activated, as a result of that, the automatic emergency shutoffs went into effect and shut the entire building and generator down. With the UPS failure, it took out quite a few systems.

In our business (as you would know), you prepare for the worst. Saying that nothing will go down and you don't need to worry about it is setting yourself up for a massive failure There is a lot of unexpected in this business that you need to be prepared for (for example, we had a quake last year in Michigan, super rare, but these things can and will affect your systems).

 

[brons2]

Don't rely too much on your UPS and your Generator, they can fail. One of our competitors had a UPS system surge and blow up, literally, causing the suppression system to be activated, as a result of that, the automatic emergency shutoffs went into effect and shut the entire building and generator down. With the UPS failure, it took out quite a few systems.

In our business (as you would know), you prepare for the worst. Saying that nothing will go down and you don't need to worry about it is setting yourself up for a massive failure There is a lot of unexpected in this business that you need to be prepared for (for example, we had a quake last year in Michigan, super rare, but these things can and will affect your systems).

Oh yea, I don't take anything for granted. I was just trying to say that, the state of write cache on my RAID cards is something that is pretty low on the scale of risk items. We'd be in a world of hurt if we had our UPS blow up. It's got a crapload of lead-acid batteries in a couple of cabinets. Looks like a cabinet full of car batteries when you open them. Where we're going is, we are consolidating 27 data centers throughout our organization into two. We are going to have 2 SAN's mirrored, 210 miles apart. Prod will be in one location and test/dev in the other. Everything 100% virtualized. Should one location go down. we will have the ability to bring up the missing server instances in the other location, since the data is mirrored. If something significant enough happens to take both locations out, there is a third DROC site in CO. Although if something like that happened the state of our data centers would probably be the least of our worries.

 

[Enlightenment]

Well power supplies can fail too, putting dangerously high voltages on a line. With professional equipment in a redundant configuration you reduce the risk of this happening, especially a BBU (for others reading: Battery Backup Unit for hardware RAID) is nice since the whole system including the disks themselves can be powered off while still regaining the buffers by the independent battery unit.

If, for example, there is a short circuit somewhere, the power supplies will cut the voltage and return to an off state, and often the power plugs need to be re-inserted before power is restored. In this case, if you have a large write-back buffer somewhere in your storage chain, you run the risk of (severe) filesystem corruption, especially NTFS and other light journaling are at risk.

About your comment with RAID5's failing: you always should have a backup. You're safer running RAID0 + a backup than just RAID5, but that should be obvious. If reliability is really a high concern, Solid State Disks should be used which should provide a 100% uptime and predictable failure (lifespan of the disk; once it hits 80% you simply replace it). And the best yet: when flash disks fail, you should still be able to read from it, so while your property is gone, your data isn't. Too bad you can't buy good 120GB flash disks for under 100 euro, but that time will come.

 

[Ockie]

Oh yea, I don't take anything for granted. I was just trying to say that, the state of write cache on my RAID cards is something that is pretty low on the scale of risk items. We'd be in a world of hurt if we had our UPS blow up. It's got a crapload of lead-acid batteries in a couple of cabinets. Looks like a cabinet full of car batteries when you open them. Where we're going is, we are consolidating 27 data centers throughout our organization into two. We are going to have 2 SAN's mirrored, 210 miles apart. Prod will be in one location and test/dev in the other. Everything 100% virtualized. Should one location go down. we will have the ability to bring up the missing server instances in the other location, since the data is mirrored. If something significant enough happens to take both locations out, there is a third DROC site in CO. Although if something like that happened the state of our data centers would probably be the least of our worries.

You outsourcing your data center hosting or are you doing everything in house?