Tweet
I use BTRFS RAID6 for my normally off home archival file server ("HASF") and I also follow the BTRFS mailing list. I am not a developer and my Linux skill set is limited so judge my comments appropriately. No one should be using BTRFS RAID5/6 in a production environment because the code is too unstable and the risk of data loss is too great. I maintain two complete backups of everything on my HASF at all times. BTRFS send |receive is a delight for incremental backup. I affirmatively try to avoid any BTRFS command which may cause any problem with the file system on this RAID6 server. Restore from backup takes time and I have experienced no problem of any kind. However, because I am not an idiot, I am somewhat of a kernel junkie and will upgrade from my current kernel, 4.0.4, to 4.1.4 and then 4.2.4 and so on as soon as they are released. I avoid the .0, .1, .2 and .3 releases because they are too risky unless they deliver specific, known functionality I need today. The development pace is fast and there will be regressions. That is the motive for holding out for the ~.4 kernel releases. When upgrading your kernel it is also suggested that you upgrade btrfs-tools to the current version using GIT. Current indications are that btrfs RAID5/6 may become suitable for expanded testing early 2016.
The mailing list contains reports of issues with RAID5/6 that may or may not cause data loss. My reading of the mailing list is that developer priority is properly focused largely on stability rather than performance at this time.
I also use BTRFS RAID1 on a petite home brouter/file server that is is rock solid stable. I am comfortable exploring the full functionality of the BTRFS file system in this environment. I stumbled across one minor regression that was easily worked around and has since been fixed. Even in a BTRFS RAID1 environment, current kernels and btrfs-tools are strongly recommended together with a verified backup strategy. I value Phoronix testing and look forward to improvements over time.
The mailing list contains reports of issues with RAID5/6 that may or may not cause data loss. My reading of the mailing list is that developer priority is properly focused largely on stability rather than performance at this time.
I also use BTRFS RAID1 on a petite home brouter/file server that is is rock solid stable. I am comfortable exploring the full functionality of the BTRFS file system in this environment. I stumbled across one minor regression that was easily worked around and has since been fixed. Even in a BTRFS RAID1 environment, current kernels and btrfs-tools are strongly recommended together with a verified backup strategy. I value Phoronix testing and look forward to improvements over time.
Comment