LisPi on Nostr: Used enterprise stuff isn't particularly dodgy and is affordable-enough one can go ...
Used enterprise stuff isn't particularly dodgy and is affordable-enough one can go for triple/quadruple redundancy reasonably (for less than the price of a single new drive of similar quality). They can be tested for mostly proper function too.
For SSDs one is plain out of luck there, testing isn't an option. Either they're spec'd for high intensity & high resilience and that hasn't been used-up yet, or they're not and there's no telling when they'll fail in some dumb way (whether new or used). The majority of them don't properly implement powerloss protection either, including the ones that claim to do so, and have the risk of corrupting even data at rest on powerloss. Which means that /regardless/ of how new a given SSD is, high redundancy is non-optional. (In theory there are commands to flush the hardware cache or disable it, but the drive can just lie about obeying it.)
Basically SSDs cause a lot of issues on this level: https://btrfs.readthedocs.io/en/latest/btrfs-man5.html#storage-model-hardware-considerations
For SSDs one is plain out of luck there, testing isn't an option. Either they're spec'd for high intensity & high resilience and that hasn't been used-up yet, or they're not and there's no telling when they'll fail in some dumb way (whether new or used). The majority of them don't properly implement powerloss protection either, including the ones that claim to do so, and have the risk of corrupting even data at rest on powerloss. Which means that /regardless/ of how new a given SSD is, high redundancy is non-optional. (In theory there are commands to flush the hardware cache or disable it, but the drive can just lie about obeying it.)
Basically SSDs cause a lot of issues on this level: https://btrfs.readthedocs.io/en/latest/btrfs-man5.html#storage-model-hardware-considerations