Raid Not Working Correctly

[Joseph Dinsdale]

Hi all,

I am currently in the proses of needing to build a new storage solution, I will be making a new topic about this later. For now I am testing out the FreeNAS software on a VM machine.

I have given the VM 5 hard drives, 1 that has FreeNAS installed and the other 4 are 4GB each, just to run a small test with.

I want to create 2 volumes. one called test1 and the other called test2. Each volume I will add 2 of the 4GB hard drives and put them in stripe raid.

im no expert in raids and what they all do but I have 3 small home NAS's and each of those are in strip raid with 2 4TB hard drives in them, so the give me 8TB each.

So from what i do know about raids is that strip raid gives increased performance and uses the storage space of each hard drive that is in the raid, that's why in my current NAS's i have 8TB storage and not less.

But using FreeNAS on this VM when I go to create a volume I add the 2 4GB hard drives and set it to Stripped raid. after the set-up of the volume, it shows as only having 4GB of space? why is this, should it not show up as having 8GB of space?

 

[jgreco]

What is the purpose of your selecting a data-protecting filesystem like ZFS if you are not going to give it any redundancy to operate from?

 

[Joseph Dinsdale]

What is the purpose of your selecting a data-protecting filesystem like ZFS if you are not going to give it any redundancy to operate from?

Hi thanks for your question. As i say i do not know much about raids i was planning on having the first volume in strip to give me the most storage then i was going to use the second volume to be a backup for the first. that's why i need the second to be in strip so it matches the same structure of what's been backed up.

 

[danb35]

By default, FreeNAS allocates 2 GB from each disk for swap space and uses the remainder for your pool. If your disks are only 4 GB to begin with, that will leave only 2 GB of usable space each.

 

[jgreco]

Well, the "R" in "RAID" stands for "redundant" and you are not providing any redundancy. When you eventually experience a disk failure, it will catastrophically wipe out the pool. And your chance of experiencing a disk failure is statistically greater the more disks you add to the stripe set.

ZFS is probably the wrong choice of storage technology if you are not interested in providing redundancy.

 

[Joseph Dinsdale]

By default, FreeNAS allocates 2 GB from each disk for swap space and uses the remainder for your pool. If your disks are only 4 GB to begin with, that will leave only 2 GB of usable space each.

Thanks for letting me know, When i go forwared with my project properly and not testing on a VM each hard drive will be 4TB, so i guess i will not notice the small 2GB take off for swap space.

 

[Joseph Dinsdale]

Well, the "R" in "RAID" stands for "redundant" and you are not providing any redundancy. When you eventually experience a disk failure, it will catastrophically wipe out the pool. And your chance of experiencing a disk failure is statistically greater the more disks you add to the stripe set.

ZFS is probably the wrong choice of storage technology if you are not interested in providing redundancy.

I am interested in redundancy very much, thats why i was going to use the second volume as a backup of the first. that way if a disk fails in the first volume i replace it and rebuild that volume, and restore the back up to it from volume 2.

 

[jgreco]

That's not redundancy. That's lunacy.

 

[Joseph Dinsdale]

That's not redundancy. That's lunacy.

why is that lunacy, i did say i dont know much about raids. is this not the best way to do this?

 

[jgreco]

The problem with your setup (including your "small home NAS's") is that you are making yourself dependent on the failure characteristics of not just one but actually both drives. If we assume that the likelihood of a single drive surviving the first two years is 90%, the likelihood of two disks both surviving is only 82%. You run about a 1-in-5 chance of totally losing the data on those devices in the first two years, and a horrifying 1-in-2 chance of totally losing the data on those devices at the 5 year mark.

This actually assumes that your devices have reasonable failure characteristics, not terrifying ones like the ST3000DM001 as reported by Backblaze (and here too). Of course, you cannot predict AFR so you have to assume that you might run into a bad batch of drives. If you get a bad batch of drives, the chance of total data loss with your scheme rises towards 100%, while the redundancy option is still down in the single digit percents, because most of the risk of data loss is that the disks would fail within the same timeframe and recovery could not be completed.

The reason your "small home NAS" devices have two bays is for redundancy. Given redundancy, they're more likely to be able to protect your data against catastrophic failure. Used the way you're using them, they're exposing your data to additional risk of catastrophic failure.

is this not the best way to do this?

Of course it isn't the best way to do that. It's basically asking fate to come and kick you in the nads. And fate will do that by failing one disk in your main pool and one disk in your backup pool before you can recover, thereby losing all your data.

You need to get over what I'm guessing is an obsession with "space maximization" and move on. A two bay NAS allows you to mirror disks for redundancy. Get your head wrapped around the fact that you will protect your data much better by using RAID-style redundancy as it was intended.

Friends don't let friends do RAID0.

 

[Joseph Dinsdale]

The problem with your setup (including your "small home NAS's") is that you are making yourself dependent on the failure characteristics of not just one but actually both drives. If we assume that the likelihood of a single drive surviving the first two years is 90%, the likelihood of two disks both surviving is only 82%. You run about a 1-in-5 chance of totally losing the data on those devices in the first two years, and a horrifying 1-in-2 chance of totally losing the data on those devices at the 5 year mark.

This actually assumes that your devices have reasonable failure characteristics, not terrifying ones like the ST3000DM001 as reported by Backblaze (and here too). Of course, you cannot predict AFR so you have to assume that you might run into a bad batch of drives. If you get a bad batch of drives, the chance of total data loss with your scheme rises towards 100%, while the redundancy option is still down in the single digit percents, because most of the risk of data loss is that the disks would fail within the same timeframe and recovery could not be completed.

The reason your "small home NAS" devices have two bays is for redundancy. Given redundancy, they're more likely to be able to protect your data against catastrophic failure. Used the way you're using them, they're exposing your data to additional risk of catastrophic failure.



Of course it isn't the best way to do that. It's basically asking fate to come and kick you in the nads. And fate will do that by failing one disk in your main pool and one disk in your backup pool before you can recover, thereby losing all your data.

You need to get over what I'm guessing is an obsession with "space maximization" and move on. A two bay NAS allows you to mirror disks for redundancy. Get your head wrapped around the fact that you will protect your data much better by using RAID-style redundancy as it was intended.

Friends don't let friends do RAID0.

So are you suggesting I set them up in Mirror raid? also this post was just about a small test I was doing on a WM but I am going to be buying s storage server that can house 16 drives. I do understand what you are saying about redundancy a bit better now. I don't want to risk my data been lost, I believed that having 2 NAS's in stripe to give me the storage space and using 1 as a backup was a good way to go but now I see that is not the case.

After I made the original post in this topic I made another post about a project I am about to start work on. If you are interested on giving me some advice on it the post can be found here --> https://forums.freenas.org/index.php?threads/advise-on-building-a-storage-solution.30634/

 

[jgreco]

The ultimate problem we face when trying to protect data is knowing how far is "far enough", and you can go quite far.

You can have lots of space at a low price and fast speed, but then you lack redundancy. This is what you've chosen. I guess it is fine if your data is not valuable to you, but it is very bad with ZFS because you've got nowhere to recover data from if there's a checksum error.

Redundancy costs you some speed, and also either increases cost or reduces space (or both). The basic unit of redundancy in ZFS is the vdev, and you can have mirrored vdevs (think: RAID1), or RAIDZ vdevs (think: RAID5 or RAID6 or beyond).

Now, the thing is, each one has benefits and drawbacks. For example, RAIDZ is very good at giving you more space for your storage dollar, and you can pick a variation with basic redundancy (RAIDZ1), or double (RAIDZ2), or triple (RAIDZ3). So if you're storing stuff on RAIDZ1, and you lose a drive, that means you've lost redundancy, and when you replace the failed drive, and it rebuilds on the new drive, you're reliant on everything on the remaining drives being readable, and not just readable, but also perfect. RAIDZ2 means you would still have redundancy, so even if you had some bad blocks on the remaining drives, you'd be fine because you could reconstruct the data on those blocks. So we usually encourage RAIDZ2 as the lowest suggested form.

The problem with RAIDZ2 is that you're losing two disks worth of storage, but that could be just two disks out of a larger number like eight. You put eight 4TB drives in the first slots of that 16 bay array and you get 24TB of usable, redundant RAIDZ2 storage. You can then later expand it with another set of eight drives. That's probably the smart option for you.

But you have to wrap your head around the idea of not being able to use every last byte of the disks that you bought. Some of the space is burned in order to get the redundancy. Your life will be easier once you've done that.

 

[cyberjock]

I just want to pipe in here and say that jgreco is totally correct. Both with what he is saying in this thread and what he said (and did) in the other thread. You need to do homework and understand this stuff for yourself. Not just so the forum doesn't ignore you, but so you don't do something silly and lose your data. Yes, there are ways that despite having backups you can totally corrupt both of them unknowingly and permanently. If you had read my noobie guide you'd already have read that you can expect to be admonished if you aren't in this to learn how this stuff works and apply that knowledge properly, not "how to make it do what i want it to".

 

[jgreco]

THIS USER HAS LEFT THE FORUM

Really, edit all your posts to say that? It destroys the context of the thread. I'm restoring the posts and locking this thread.