I have a pool that was originally set up with a mirror of 2 HDs each 2TB. One of the disks failed and I replaced it with 8TB disk. The way I replaced it was:
Mark the bad disk as offline.
Physically attached the new 8TB disk
Assign the 8TB disk to the existing pool
The pool is still 2TB in capacity after resilvering.
Now I just bought another 8TB disk hoping to replace the remaining 2TB disk, with the end result of an 8TB pool utilizing all the disk space on the 2 mirrored drives.
The questions I have:
There is a “Replace” button on the UI. Should I use that instead of the “detach” etc steps?
There is also an “Expand” button next to my existing pool. Does it have any role in my end result, e.g. should I first “replace” 2TB disk with 8TB, then click “Expand”, or does Truenas scale do the expanding automatically?
Based on other people’s comments previously, I think Truenas should automatically increase the capacity of the pool once I replace the 2TB disk with 8TB, but I am unsure what does the “Expand” button do.
The “Expand” button is sort of a relic of the past, before the default “autoexpand” feature was set on pools.
Don’t use “detach” for this.[1] Either “offline” → “replace” (if you require a reboot and physical swapping of drives), or just skip to “replace” if you have the available slots in your case to house all the drives.
OK. I just replaced the 1.5TB disk with 8TB.
I first marked the 1.5TB disk as offline.
Powered off the machine, and connected the cables to the 8TB new disk.
Then start TrueNas, and clicked “Replace” on the offline disk, and selected the new 8TB.
After the resilvering, the pool still shows 1.5TB capacity.
There does not seem to be an automatic increase of capacity after.
Did I do something wrong? Do I click Expand button now to get the full 8TB capacity?
autoexpand has been set by default since forever, but apparently iX has forgotten how to partition a disk to use all its space when doing a replacement. This bug has persisted through several versions of SCALE now, and they’ve (falsely) said it’s been fixed at least a few times. I thought it was fixed in 24.10, but perhaps not.