Ashift value for SSD WD Red SA500

TrueNAS General
,
1

Hi there, my first question in the community here :slight_smile:

I am finishing setup my NAS. It has a main raidz2 pool with 5 Seagate IronWolf 4T each. I was happy a month ago, when I finally got to finish it, and then started reading about the special vdev and possible small files usage there too.

To make it short, I decided it was a nice thing to try, but it would be nice to get it right from the first go. So I put it on hold and ordered 3 ssd WD red 1T to create a special vdev for this. Today they arrived.

Doing a smarctl to them I see it informs:

Sector Size: 512 bytes logical/physical

I’ve read a lot about it, but I am not sure if that is safe to assume that is the ultimate true, or if it has a different distribution under the hood. All this just to set a proper ashift value now that I am creating the pool.

I’ll appreciate your experiences and opinion on this. My inclination is towards a value of 12 anyways, but I’ll wait a bit more for possible feedback.

As usual, thanks a lot!

2

You do not get to choose the ashift when adding a vdev, it will be the same value as that already assigned to the pool—which should be 12.

Think hard whether you really want a svdev and what you expect from it. 5-wide Z2 is rather small, and I suppose that data is already in the pool.

3

oh, that was silly on me then. 12 then it’ll be.

About the planning I did the readings for months, and without being 100% sure went for the raidz2. The problem with the simpler mirrors is that if you get unlucky and 2 disks go away in the same vdev that is it. And mirroring in 3 ways or more is too expensive. With this, my idea was to have at least room for 2 of them to fail. anyone, and yet still be able to try resilvering. I know it will stress big time when the time comes, but it is what it is. The feeling of security I get with this,. seems better to me, but it is just me maybe. Going further on disks, equipment etc is out of reach for me too.

With the SSD addition I plain to gain just a bit of performance, maybe, but I really don’t care. The main propose here is resilience. But when I realised than setting the record size equals to special small block in a dataset for instance will make use of the SSDs only, was the plus I wanted just in case it comes handy for some documents and minor stuff in the future. I understand the risk of the metadata here too. But tried to match the same level of redundance.

Went through several good readings, among them this and this

Anyways, thanks for the clarification!