Been thinking about this odd RAID setup scenario and wanted to get some input from folks here. Say you're putting together a RAIDZ-2 array with 8 disks total, four 10TB ones and four 5TB ones. The goal is to handle failures without losing data, and maybe squeeze out more space.
Here's the breakdown:
First option: Treat everything as 5TB disks, so you end up with an effective 8 x 5TB array. That means you could lose up to two disks and still be okay.
Second option: Combine those 5TB drives into pairs using RAID 0 to make them act like 10TB drives, then build a 6 x 10TB RAIDZ-2 array. If a RAID 0 pair fails, it's like losing a whole drive, so you'd need three of those fake 10TB drives to go or a mix like one RAID 0 failure plus two real 10TB ones. Weirdly, this seems to survive more specific failure combos than the first setup, even though it uses RAID 0.
I did some quick math, and it points to the second option being tougher in certain cases, which doesn't sit right with the usual advice. Like, we all know RAID 0 is risky, right? But here it might give better space and handle an extra failure or two if you're lucky.
What's the real story here? Has anyone tried something like this in their homelab? Any pitfalls I'm missing, or better ways to handle mixed drives? Maybe stick with all the same size drives, or use some other tricks for redundancy? Love to hear your setups and thoughts on making this safer or more efficient.
