Replacing old cache SSDs in Unraid: what worked for me

Replacing two old cache SSDs in Unraid (what worked for me)

Why I did this

My cache pool was running on two Samsung 850 EVO 500GB SSDs that were close to 10 years old. They were still working, but given their age I wanted to replace them proactively before a failure, while keeping the cache pool and data intact.

My setup

• Unraid 7.2.3

• Cache pool: Btrfs, RAID1

• Old drives: 2 × Samsung 850 EVO 500GB

• New drives: 2 × Samsung 870 EVO 500GB

• SATA hot-plug enabled in BIOS

I did not shut the server down at any point during this process. All drive swaps were done live.

Before starting (important)
Before touching any disks, I ran an Appdata Backup using CA Backup/Restore. Nothing went wrong, but having a current backup made the whole process much less stressful.

What I expected (and what didn’t work)

I expected to be able to hot-swap each old SSD with a new one using Unraid’s Replace function. Instead, when I tried to replace the first drive, Unraid failed with an error like:

Wrong Pool State – replacement device too small

Even though both drives were advertised as 500GB, Unraid compares usable size, not marketing size, and the replacement was rejected.

What actually worked (important part)

This is not a true “replace”. It’s a remove → re-add process.

I did this one drive at a time.

First drive

1. Stop the array

2. Physically remove one old cache SSD (hot-swap, server stayed on)

3. Start the array
   Unraid reports a missing cache pool device
   Check the box to remove the missing pool disk

4. Stop the array again

5. Set Cache Slots = 2
   Unraid drops this to 1 automatically, which is easy to miss

6. Assign the new SSD to the empty cache slot

7. Start the array

8. Let Btrfs balance run (it started automatically for me – do not interrupt it)

9. Verify RAID1 with:
   btrfs filesystem df /mnt/cache

Do not proceed until Data, Metadata, and System all show RAID1.

Second drive

Once RAID1 was fully restored, I repeated the exact same steps for the second old SSD.

How I verified success
After each replacement, I ran:

btrfs filesystem df /mnt/cache

The final, correct state for my setup was:

Data, RAID1: total=69.00GiB, used=67.36GiB
System, RAID1: total=32.00MiB, used=16.00KiB
Metadata, RAID1: total=2.00GiB, used=375.28MiB
GlobalReserve, single: total=120.09MiB, used=0.00B

The Unraid UI showed “Pool of two devices” with both new SSDs.

Key takeaways

• Same advertised size does not guarantee a cache replacement will work

• If you see “replacement device too small”, remove the old device first, then add the new one

• Do not start the second replacement until btrfs filesystem df /mnt/cache shows Data, Metadata, and System as RAID1.

• Watch the cache Slots setting – it will trip you up if you miss it

• Let Btrfs balance finish before doing anything else

• If hot-plug is enabled, this can be done without shutting the server down

This was done after I did this so my instructions may not be perfect but hopefully they help...Thanks!

Edited February 7Feb 7 by TexasDave
Cleaned up formatting