Replace all drives and some hardware

[gilgedje]

Hi everyone!

im runing Truenas Scale 22.12.3.3
CPU: CPU E5-2680 v3
RAM: 256GB ECC samsung ram
Storage: 10 Toshiba MG04SCA60EA 6TB sas drives (12G)
Network: 10Gig
I have 1 pool with 2 raid z1
The pool status is OK! just finished a scrub yesterday - no errors on any disk.

I bought new drives - 17 of the Segate Exos X16 10TB (all new)
I want to change the entire setup - i even bought a new motherboard (supermicro x10dri - using 2 of the 2680 v3)

The question is - what should i do about replacing the drives?
Should i build the new server containing 3 raid z1 (5 drives each) and move the data?
Or should i replace the current disks and expand the pool and then replace the hardware?

If its important in any way - i used to change the current disks (i had 5 of the 4tb WD red plus drives, and i changed them the toshiba drives and added 5 more to have the current setup)

 

[MrGuvernment]

What are you using the system for and why Raid Z1?

 

[gilgedje]

What are you using the system for and why Raid Z1?

Mostly media for plex
Raid z1 because when i started i had 5 drives (as i mentioned above)
So i wanted to get the most storage and i had small drives (4tb) with spares to change any faulty
When i changed to the 6tb drives i wanted to keep the media so i added a raid z1 with 5 6tb disks to the same pool and changed the original raid z1 to the 6 tb drives - getting 10 6tb drives

My question was how to proceed knowing its not the best configuration for my future system and since i already change almost everything i can move the data to a diffrent configuration server

 

[ChrisRJ]

I want to change the entire setup

How? Why? What is your use-case? What aspect does the current setup not address properly?

 

[gilgedje]

How? Why? What is your use-case? What aspect does the current setup not address properly?

Today i am able to recreate the entire server is what i mean
I have a new chasis, new disks and hardware.

Since i change the disks to a higher capacity disks my main concern is losing data while changing faulty drives in the future since it will take longer to replace a faulty disk

And my main question is how should i build the new server?
Should i just replace the disk and stay in the same configuration cause having 3 raid z1 is somewhat of a 3 disk redundant
Or should i build a single raid z2 with everything i have - although i understand the iops will decrease too

 

[Patrick M. Hausen]

A single 17 disk wide vdev is not recommended. I would go for 2x 8 RAIDZ2 and one hotspare.

 

[gilgedje]

A single 17 disk wide vdev is not recommended. I would go for 2x 8 RAIDZ2 and one hotspare.

I have 17 disks but im looking to put 15 of them in the server keeping 2 aside for a bad day
In that case, what would you do?

 

[ChrisRJ]

@gilgedje , it would be helpful to know what your use-case is. Without telling people what you want to do with your NAS, can only give generalized information. But if you want to do video editing, that requires a totally different setup compared to archiving pictures.

 

[gilgedje]

@gilgedje , it would be helpful to know what your use-case is. Without telling people what you want to do with your NAS, can only give generalized information. But if you want to do video editing, that requires a totally different setup compared to archiving pictures.

As i said, mostly plex media
Its most of the use case
I have arrs and qbit to download media
I have a gpu for transcoding and i stream to multiple people - maximum of 10 people
I also have about 2 tb of ssd nvme for cache

So i mostly stream media.

 

[Patrick M. Hausen]

2x 7 plus hotspare or 3x 5 RAIDZ2 ... I am very conservative with my data and would never use RAIDZ1 for long term storage. And you should not go over 12 disk wide vdevs. Then - just pondering - with that number of disks maybe DRAID is for you. That would imply pool creation on the command line but TN can perfectly well use it afterwards.

 

[Etorix]

As i said, mostly plex media

Then raidz# is adequate, save for downloading torrents, whose constant write activity with small blocks would do better on mirrors (or even a single drive…).
Raidz1 is advised against with large drives, for data safety in the advent of failure. So 2 * 8-wide raidz2 + 1 spare (hot or cold) is really optimal. You may consider 2 * 7- wide raidz2 + 3 spares, or raidz3 if you're worried. 3 * 5-wide raidz2 + 3 spares is possible, but looks to be excessively cautious and not very efficient. Anything involving raidz1 is a lot less secure than 2*8z2+1s.

I also have about 2 tb of ssd nvme for cache

What do you mean by "cache"?
With 256 GB RAM you can actually afford 2 TB L2ARC without performance penalty… but it's likely that you get no benefit out of it because ARC is already large enough.
If this is a SLOG, you have no stated use for it, and I bet that the drive is not suited to this application.