Bit Rot Prevention Scheme

While this probably doesn’t hurt, it should normally not be necessary as ZFS doesn’t return this virtual folder in a readdir-style system call, so syncthing won’t pick it up during a scan anyway.

(Unless snapdir is set to visible)

1 Like

Hmm I was unaware of visibility control, indeed snapdir is a parameter which can be controlled. Thanks for the tip! Learning every day.

2 Likes

Good strategy overall.

I am personally uncomfortable with copies=2 on two drives. I’m less worried about on-disk loss than I am about full-disk loss.

My personal “minimum equipment list” is data on 3 disks:

If one disk dies, the second disk now contains your only copy of the data, and will now be subjected to a severe strain when re-silvering (or duplicating) the data to a replacement disk. IF it is going to die, re-silvering is the most likely time for it to happen.

If you have a mirror with an off-line backup drive, you’re still in the same state with only one drive containing all your data: your second mirror drive dies while re-silvering, you fall back to your offline drive, and it dies. Highly unlikely, but I’m unwilling to make that my primary plan.

Yeah you still need an extra backup. I do this on USB SSD/HDD once in a while.

FWIW: My minimum is always RAID-Z2 with > 4 physical disks and a disaster-recovery drive in another physical location (in-laws house) that gets updated every month or so using zrep to send ZFS snapshots to the backup drive.

zrep rolls up intermediate snapshots from the host system so I have the complete filesystem history, and since ZFS knows what blocks have changed between snapshots it doesn’t need to “walk the tree” to copy the changes and backup updates happen pretty fast.

1 Like