RAID5 drives or JBOD?

[myblade]

Hi,

 

I am about to extending my pool with a new set of drives.

 

In the past I moved from drive bender to this because 1 drive died / got stuck and I wanted to take advantage of the Scanner.

 

So my current setup is a bunch of single disks and a RAID5 pool to prevent data loss of one drive, all added to one pool.

 

I am now asking myself if it is still clever with drivepool and the scanner in place to build a second RAID5 drive with the new drives and add it to the pool or to keep it as a bunch of disks and rely on the scanner?

 

Is there any advantage / disadvantage in using RAID5 for pooled drives or leaving them as a bunch of disks?

 

I don't want to duplicate stuff, because this would lead to heavy requirements. At the moment I am at about 20TB for my media library and I don't want to double the required space With the new disks I would now be able to move the data off the RAID and reset them as a bunch of disks. This is why I am asking.

 

Thank you in advance.

 

Kind regs

[Christopher (Drashna)]

Well, this is a fairly loaded topic. 

 

 

You may want to read this:

http://community.covecube.com/index.php?/topic/52-faq-parity-and-duplication-and-drivepool/

And this:

http://jrs-s.net/2015/02/06/zfs-you-should-use-mirror-vdevs-not-raidz/

(while this is for ZFS, a linux file system/RAID solution, it still applies here).

 

But basically, If you're using hardware RAID, you may be giving up the ability to move the drives to different hardware, especially if that controller dies. 

 

If you're using storage spaces to do this, then you may be giving up the ability to recover any data if something catastrophic occurs, as there are very few utilities or places that do data recovery for Storage Spaces. 

 

 

 

As for the not wanting to use double the space for duplication, I absolutely understand that. It gets expensive fast, if you have a large amount of data. 

 

However, keep in mind, with a larger RAID array, rebuild times are going to be in the days or weeks, depending on the hardware involved. During that time, you will experience a significant drop in performance. And depending on the exact circumstances, you may be at an increased risk of catastrophic failure (if all the drives were purchased at the same time and are from the same lot, they will likely have the same defects, and the extra load on the drives may trigger the same sort of failure) 

 

 

 

Additionally, depending on exactly how you're setting up the RAIDs, StableBit Scanner will be unable to read the SMART data from the underlying drives, essentially giving up a large chunk of diagnostic info. 

That said, the monthly surface scans and file system checks is still absolutely helpful, even in on a RAID. 

 

 

And if you were not aware, if StableBit Scanner does detect unreadable sectors during the surface scan,  it will cause StableBit DrivePool to evacuate the contents of the drive, to help prevent data loss. 

 

For a large RAID5 array, it means that there may not be enough space on the rest of the pool to accomplish this, and if the array catastrophically fails, you will most likely lose data anyways. 

 

 

 

 

If you really want parity protection, the first link includes a number of user threads talking about how to set up SnapRAID with StableBit Drivepool.  SnapRAID is a snapshot parity software package that lets you specify a single disk as the parity disk and allows you to protect regular NTFS volumes.  This may be a better solution for what you want, as it works entirely within Windows, allows full access to all of the drives (in the case of StableBit Scanner), and should be 100% portable. 

[myblade]

Thank you for your post.

 

I did not mean the general information on RAID and stuff itself, but in the concrete use here in drivepool I've already read the other thread, but was unsure on applying it to my situation.

 

But I understand the loss of functions when RAID is used.

 

So the Scanner can not make use of the SMART information. And in case it should evacuate the drive this would effectively concern the whole array and thus leads to more free space required.

 

My current situation is:

 

- 1x 12 TB RAID5-Array

- 2x Single 2TB disks

- 2x Single 3TB disks

 

My plan was/is:

- Add a the new disks as a RAID5 array

- Add 1 additional disks each size of 2TB/3TB to build (software) RAID5 arrays with the single disks

-> Pool consisting of 4 RAID5-Arrays.

 

Why this was my plan?

I don't wanna use duplication, although this was already available in drive bender. The needed drives for doubling the space are far too expensive.

So there is no security, but to rely on the scanner, when using single disks without RAID.

 

In the past one drive suddenly died. There was no SMART notification or anything announcing the potential loss of my disk. Since I do not seem to be able to get the data of the disk again, this is what I want to avoid now.

 

I see the Scanner as an additional help checking the disk, but with RAID-Arrays added to the pool, an automated evasion is hardly possible. This would need about 66-75% of the size of the biggest array in the pool as free space across the pool.

 

If my enclosure dies I could replace it with a similar one.

If a drive dies, I can swap it out. It surely takes its time to rebuild, but its better than the complete loss if the drive went fully broken in a sudden.

 

Määäh... it seems that you have to pick your poison

 

Thank you anyway

[Spider99]

an added thought - raid5 having only one level of redundancy - when you change a failed drive the others can fail during the rebuild process - then you loose all your data on that array - this is one of the reasons i have gone away from raid that i have used for years to DP - also depending how you create the arrays they are not usually easy to move to another machine and work out of the box as DP will.

 

Sounds like you only have 3 disk raid5's or plan to - in that case you have 33% redundancy and its not that far off 50% with DP

[Christopher (Drashna)]

In a RAID array, yes, StableBit Scanner won't be able to grab SMART data.  

 

However, the surface scans still work and can detect issues on a RAID array, without any issues.  
And yes, if a bad sector is discovered there, it would cause the pool to attempt to evacuate that "drive". 

 

 

As for the disk suddenly dying, that can happen, even when SMART data is available. There are a number of reasons for this, but a defect in the PCB or the chips on it could cause this. As could a power spike/short. 

 

 

As for the no redundancy, I used to do this as well. Mostly for the videos, as I have a LOT, and I suffered multiple disks failures in a short period of time.   

To be blunt, it's easier to spend the money on more disks, and duplicate, than it is to attempt to rebuild, in this case.  

 

But to be fair, I'd recommend using SnapRAID over hardware RAID.  There are a number of reasons, but the main is that you can specify a single disk for the parity disk.  You could even use a mirrored drive, so you don't have to worry about it failing. 

The advantage here, is that the rest of the disks in the system would not be in any sort of array, but handled by the OS.  
 

This would allow you to not have to use duplication/mirroring for the data, but allow you to protect a data in a way that would maximize the use of StableBit DrivePool and StableBit Scanner. 

 

The caveat is that the parity disks would have to be the largest disk(s) in the system.  

[myblade]

an added thought - raid5 having only one level of redundancy - when you change a failed drive the others can fail during the rebuild process - then you loose all your data on that array - this is one of the reasons i have gone away from raid that i have used for years to DP - also depending how you create the arrays they are not usually easy to move to another machine and work out of the box as DP will.

 

Sounds like you only have 3 disk raid5's or plan to - in that case you have 33% redundancy and its not that far off 50% with DP

 

Yep, I know, there is only security for one failed drive. This is why I already thought about using Storage Spaces for Dual Parity, but this would drop the whole feature set of DP.

 

In my whole life I have suffered only 2 failed drives. One due to a head crash caused by a speaker dropping on my pc while it was working and the other one I mentioned. I may have been lucky but the one 

 

The existing array consists of 4 disks and the new would also consists of 4 drives. This is due to the limitations of the enclosures.

 

The single disks are internal and these arrays could consist of 3 drives each.

 

In a RAID array, yes, StableBit Scanner won't be able to grab SMART data.  

 

However, the surface scans still work and can detect issues on a RAID array, without any issues.  

And yes, if a bad sector is discovered there, it would cause the pool to attempt to evacuate that "drive". 

 

 

As for the disk suddenly dying, that can happen, even when SMART data is available. There are a number of reasons for this, but a defect in the PCB or the chips on it could cause this. As could a power spike/short. 

 

 

As for the no redundancy, I used to do this as well. Mostly for the videos, as I have a LOT, and I suffered multiple disks failures in a short period of time.   

To be blunt, it's easier to spend the money on more disks, and duplicate, than it is to attempt to rebuild, in this case.  

 

But to be fair, I'd recommend using SnapRAID over hardware RAID.  There are a number of reasons, but the main is that you can specify a single disk for the parity disk.  You could even use a mirrored drive, so you don't have to worry about it failing. 

The advantage here, is that the rest of the disks in the system would not be in any sort of array, but handled by the OS.  

 

This would allow you to not have to use duplication/mirroring for the data, but allow you to protect a data in a way that would maximize the use of StableBit DrivePool and StableBit Scanner. 

 

The caveat is that the parity disks would have to be the largest disk(s) in the system.  

 

Yep. Of course this can happen as it just did. It was one of my single drives and even drive bender does check SMART data, but it just died right away.

 

Also any other redundancy solution is not safer than a kind of backup solution (like duplication).

 

I just had a quick view on the page of snapRAID and it seems to keep the disks as they are, like DP and DB. They keep the NTFS as the file system and just add some features like parity on file system and application level, right?

 

Maybe I should have a closer look. This may be the solution in combination with DP I seem to be looking for. But if it is so, I have to change the enclosure Then there is a better one with UASP for single drives