More Parity

One parity protects against one failure at a time. SnapRAID supports up to six parity levels, each an independent computation over the same stripes:

Config line	Name in reports	Failures survived
parity	Parity (P)	1
2-parity	2-Parity (Q)	2
3-parity … 6-parity	R, S, T, Z	3 … 6

"Failures" counts in units of damage: two dead disks, or a dead disk plus a scattered corruption elsewhere, or two files' worth of bad blocks in the wrong places. The lab ran one and two levels; three and above are the same mechanism with more files, and are documented-only in this guide — the manual covers their configuration, which differs only in the extra lines.

What an extra level costs and buys

Each level costs one more disk (or the space of the largest data disk on one), and more computation at sync time — parity maths is fast, and on any modern CPU the disks are the bottleneck, but the parity files themselves are real terabytes on a real array.

What it buys is not mainly "two disks die at once", which is rare. It buys surviving the unlucky overlap: a disk dies, and during the long rebuild another disk turns out to have a rotten patch — the classic way single-parity arrays lose data, because the rebuild reads every surviving block and finds the ones that were quietly bad. If your array is large, your files irreplaceable, or your rebuild window long, the second parity is cheap insurance aimed at exactly that day.

Adding a second parity to a running array

The lab did this for real (transcript t11), and the procedure has one non-obvious step: a plain sync refuses when it finds a new parity file shorter than the array expects — quite rightly, because "parity disk not mounted" looks exactly the same. It tells you so:

DANGER! One or more the parity files are smaller than expected!
It's possible that the parity disks are not mounted.
If instead you are adding a new 2-Parity level, initialize it with
'snapraid -d 2-parity fix', and then run 'snapraid sync'.
Alternatively, use 'snapraid --force-full sync' to rebuild all parity levels.

The lab took the second offered path:

$ snapraid -c snapraid-dual.conf --force-full sync
...
Everything OK

--force-full recomputes every parity from the data — the expensive, certain option. Afterwards both parity files stood at the same size, and the test: corrupt a file on d1 and wipe d2 at the same time. fix recovered everything; every checksum verified:

recovered photos/holiday-snaps.bin
recovered movies/birthday.mp4.bin
recovered movies/captions.txt
recovered music/concert.flac.bin
recovered music/new-track.flac.bin
       0 unrecoverable errors

Split parity, briefly

A parity level can be split across several smaller disks when no single disk is big enough — parity accepts multiple paths in sequence as the file grows, and the manual documents the syntax. The lab did not run a split; it changes capacity planning, not concepts, and the refusal messages in chapter 14 mention split files because SnapRAID guards each piece the same way.

How many levels, honestly

• One — the home media array, files also backed up or re-obtainable. The common case, and a reasonable one.
• Two — large arrays, long rebuilds, or data you cannot get back another way.
• Three or more — specialist territory; at that point also ask whether a second copy of the data (a backup, chapter 13) is the better spend, because parity of any depth still has no history and no answer to deletion plus sync.

The one-sentence version

Each extra parity level costs a disk's worth of space and survives one more simultaneous failure — and the second level, aimed at damage discovered during a rebuild, is the one most home arrays that add any, add.