Cache and BTRFS

[brainbone]

There seems to be a general consensus here that btrfs doesn't handle power loss well.   Seems people often loose much of the data on their cache drive(s) if power is cut to the system.

 

Even though I'll be running with a UPS, from my point of view, total data loss of the "cache" (which is used for far more than just cache in unraid) from a simple power failure is simply unacceptable.  It is not uncommon at all for a UPS to fail just at the moment you need it.

 

What can be done to make the cache drives more resilient in the case of power loss?  Can the cache be run with XFS or EXT4?

 

Edited August 15, 2018 by brainbone

[phbigred]

As far as I know nothing says cache needs to be btrfs. I think you can format it as xfs as well, just lose the pooling feature. 

[Chris Pollard]

Hardware raid of some sort? I'm not a fan of BTRFS either although I have managed to recover things after failures.

[JorgeB]

In theory btrfs should be better than xfs with unclean shutdowns, in practice it's not, at least no yet, but if there's corruption most times all or almost all data can be recovered with an ro mount or with btrfs restore, still if you don't need a pool and don't care for checksums or snapshots you can use xfs instead.

[brainbone]

Thanks, all, for confirming that I can use xfs on the cache.

 

I do like the idea of btrfs, but the apparent reality scares me a bit.

 

Thinking more about the issues with unclean shutdowns, is there any known reason why this could be happening?  Is unraid using btrfs without barriers?  Does this issue only occur on drives without power loss protection?  etc.

 

[pwm]

Btrfs should be very good at handling power loss for single-disk use.

 

The open question is how it handles different variants of disk pools, where there are cross-disk synchronization issues introduced.