Drive upgrade path?

[pitt1717]

so i just bought 4 new drives (3tb each). I'm looking to use 3 for storage and the 1 for parity. 9tb usage

anyway i currently have a 3 drive setup (1tb) each. 2 for storage and 1 for parity. 2tb usage

i know (or i think i know) i can upgrade each disk 1 at a time from the 1tb drives to the 3tb drives. but can i add a disk at this time as well? or do i need to create a new raid altogether?

ie
1tb a,b,c
3tb 1,2,3,4

replace a with 1, b with 2, c with 3, and add 4

 

[danb35]

You will need to create a new pool. Be aware that RAIDZ1 (only one disk of parity) doesn't protect your data as well as you might think--see cyberjock's sticky and the links in his sig for more information.

 

[pitt1717]

i figured i would need a new pool. i would love to do Z2 or 3 but i can't stretch the funds for 2 more 3tb drives. also the 3 main things (music, photos, videos) are stored on my laptop, desktop and NAS. so i always have a backup. the NAS drive upgrade is due to my dvd collection, and as i have the disks the loss will be more of the time lost than anything. but that pdf was enlightening and will at least try to get some better monitoring in place

 

[Ericloewe]

a) RAIDZ1 is a bad idea if you value your data.
b) You cannot grow vdevs. You need to destroy and start over.
c) 4 disks is bad for RAIDZ1

For more information on c), here's a bit of canned information you might find useful:


Recommended vdev sizes are 2^n+p, with n={1, 2, 3} (larger vdevs are possible but definitely not recommended) and p the number of parity drives (The RAIDZ level)

This means:
RAIDZ1 (which is a bad idea if you value your data): 3, 5 or 9 disks
RAIDZ2: 4, 6 or 10 disks
RAIDZ3: 5, 7 or 11 disks

Other configurations lead to misaligned ZFS sectors (kinda like the 4k drive issue a few years ago), which leads to performance issues. It may also lead to significant losses of capacity (even if it doesn't, fragmentation will be noticeable sooner than with optimum configurations).

I suggest you read Cyberjock's guide - it'll answer your basic questions.
http://forums.freenas.org/index.php...ning-vdev-zpool-zil-and-l2arc-for-noobs.7775/

 

[joeschmuck]

@pitt1717

A few comments:
1) As indicated above you cannot upgrade your current pool from 3 drives to 4 drives, if you did and the new 4th drive failed, all would be lost.

2) You can choose RAIDZ1 or RAIDZ2, it will depend on how redundant you need your data to be. As you described above, you have the data stored elsewhere so I don't see a problem with RAIDZ1.

3) The optimum configuration for a RAIDZ level does depend on specific numbers of drives however lets be clear here, if you are not creating a high volume/high transaction (like a search engine for example) then you don't need "optimum" configuration. Very few people do have this configuration for a home system because it's irrelevant to stream videos, store backups, etc...

a) RAIDZ1 is a bad idea if you value your data.
b) You cannot grow vdevs. You need to destroy and start over.
c) 4 disks is bad for RAIDZ1

For more information on c), here's a bit of canned information you might find useful:


Recommended vdev sizes are 2^n+p, with n={1, 2, 3} (larger vdevs are possible but definitely not recommended) and p the number of parity drives (The RAIDZ level)

This means:
RAIDZ1 (which is a bad idea if you value your data): 3, 5 or 9 disks
RAIDZ2: 4, 6 or 10 disks
RAIDZ3: 5, 7 or 11 disks

Other configurations lead to misaligned ZFS sectors (kinda like the 4k drive issue a few years ago), which leads to performance issues. It may also lead to significant losses of capacity (even if it doesn't, fragmentation will be noticeable sooner than with optimum configurations).

I suggest you read Cyberjock's guide - it'll answer your basic questions.
http://forums.freenas.org/index.php...ning-vdev-zpool-zil-and-l2arc-for-noobs.7775/

Eric, I know you mean well but you should temper your answers to the person asking the question. I don't agree that a home system for this person, and with all the backups they already have needs more than a RAIDZ1 if they don't want it, nor an optimum drive setup. The alignment issue you speak of doesn't lend itself to significant anything and fragmentation is part of ZFS, or in other words, ZFS is all fragmentation, there is no data which is contiguous. Or maybe I misunderstood you, which is entirely possible.

So, I'm not meaning to step on any toes, I do my best to keep the peace and hope my information is helpful.

 

[Ericloewe]

@pitt1717
A few comments:
1) As indicated above you cannot upgrade your current pool from 3 drives to 4 drives, if you did and the new 4th drive failed, all would be lost.

2) You can choose RAIDZ1 or RAIDZ2, it will depend on how redundant you need your data to be. As you described above, you have the data stored elsewhere so I don't see a problem with RAIDZ1.

3) The optimum configuration for a RAIDZ level does depend on specific numbers of drives however lets be clear here, if you are not creating a high volume/high transaction (like a search engine for example) then you don't need "optimum" configuration. Very few people do have this configuration for a home system because it's irrelevant to stream videos, store backups, etc...



Eric, I know you mean well but you should temper your answers to the person asking the question. I don't agree that a home system for this person, and with all the backups they already have needs more than a RAIDZ1 if they don't want it, nor an optimum drive setup. The alignment issue you speak of doesn't lend itself to significant anything and fragmentation is part of ZFS, or in other words, ZFS is all fragmentation, there is no data which is contiguous. Or maybe I misunderstood you, which is entirely possible.

So, I'm not meaning to step on any toes, I do my best to keep the peace and hope my information is helpful.

You're right about the extra backups, my response was based on the original post, the new one appeared while I was writing.
In light of the presence of backups, I do agree that RAIDZ1 is an acceptable compromise, with all the obvious strings attached.

Regarding the alignment: As insignificant as it sounds, it means the drives end up having to access a lot more sectors to access a ZFS block, since there are now unfilled sectors. What I mean is that as long as there's enough room for everything, the performance drop stays within the acceptable range. Once space starts drying up, ZFS will have to either discard the empty sector pieces, reducing available space, or start filling the unfilled sectors, which means very bad seek times and associated performance penalties.

Don't worry about disagreeing, it's fine ;).

 

[joeschmuck]

I had no idea the OP was updated. But just so you know, I do promote RAIDZ2 most of the time since I've walked down that road before. One thing we never found out was the type of system the OP has. If I found out it has Non-ECC RAM or very low RAM then I'd be in favor of UFS. I am trying to answer only the question put up in front of me since users are not following the forum rules and they are not posting their hardware and FreeNAS version quite often. And I'm tired of asking for it all the time, it just makes you look like a snob or unfriendly, even though it really is the right thing to do.

 

[pitt1717]

Guys. Please don't argue. I am taking your advice. Reading that pff and going to make a better strategic decision as I learn a bit more.
@joe your thinking is why I went with 4 drives. Looking for 9tb storage with the ability to limp with 1 failure while I get a new driv (<1week). This is for mainly streaming and a second pool on the nas to backup pics and music.
I really don't need the redundancy for ISO's etc
But again no need to argue as I appreciate any and all advice


Sent from my iPhone using Tapatalk

 

[anodos]

no need to argue as I appreciate any and all advice

In that case you should just stripe the drives. No need to worry about whether you will have a URE or second or third failure during resilvering. You can get straight to the business of restoring from backups. :)