sata* to /drv/ad* mapping

[eyocum]

Sorry about posting here, but I'm not sure how to ask the question, and my searches haven't turned up an answer - please point out where I can find the information.

I'm currently testing/playing with drive replacement, and I've run into something odd (I thought). I was trying out the change drive process, when I saw that the controllers were not being mapped to the same ada number that they were physically cabled to (sata0 cabled to ada0 cage, sata1 to ada1 cage, etc). After I replaced a drive and completed the rebuild process, I saw that the new drive was in ada2 per the system, even though I had replaced the drive that was ada1. I removed the new drive, and the current labeling remained (1,3 and 4 populated, 2 empty, but physically sata 1 is empty, and 2,3,4 populated).

What happened? I have more questions, but they'll probably be answered by the right source material...

Thanks for guidance,
Erik

 

[sretalla]

Don't pay attention to the assignments like DA0 or ADA0 or whatever... FreeNAS/ZFS uses GUIDs (actually GPTIDs) to identify disks which are written to the disks themselves, so they will work no matter which port/controller they are connected to.

 

[eyocum]

Ok, thanks for that. How do I know which physical drive is mapped to which ad* after a change? Do the descriptions stay static (not using them yet, I was bit by this), to the drive even if the drive (ad*) names change? i.e. if ada1 is Desc "Drive 1", the field stays even if the ada1 label goes to ada3 because of the system managing the disks via GPTID?

 

[danb35]

How do I know which physical drive is mapped to which ad* after a change?

By serial number.

 

[eyocum]

Yes, I was just hoping the 'Description' field would stay static, less labeling of the case to change if I had to change a drive :)

 

[danb35]

I think it should stay static, but I can't guarantee that it will.

 

[Ericloewe]

Device IDs are supposed to be statically tied to the physical ports, but numbering is not likely to match. Don't trust that, though.