Amen to that. Also, I’m really unimpressed with (or really bad at using) RAID 1 on Linux…it’s not clear to me at all if there is any way to detect which copy got corrupted…
I have had my coinstats index on my satellite node get corrupted but no errors on fsck…
Backup probably needed in addition to dual disk redundancy…
Backup doesn't work for the LN state database. You need real-time copies to be stored redundantly or you're screwed.
You want RAID-1 level redundancy but block-level mirroring isn't the modern way to do things anymore for exactly the reason you state. Filesystems like btrfs use checksums so silent corruption doesn't happen anymore. You can specify which underlying partitions to use at RAID-1 level. The disks don't even need to be the same size. You could have one big drive to store everything including the full node while a small disk is specified to provide raid-1 redundancy for only the LN datadir.
I don't personally do it that way because I want both of my disks to be bootable to make it easier to recover from failure. However this big+little disk approach reduce cost of a redundant storage LN node.
I’ll have to try btrfs…I treat .bitcoin/ like a home directory and keep it separate from my system disk
Btrfs lets you set per directory raid-1 even after the filesystem is mounted. Please read the documentation carefully and maybe even practice in a virtual machine with big+small virtual disks.
VM is a good idea. Part of why I haven’t run a lightning node is me not being good at data redundancy/real-time backup…I thought I was clever with raid1, right up until I had a discrepancy and no way to know which was right automagically…
