Thus if you don't make metadata raid 5 or 6, you avoid that problem. Simply put the 'write hole' bug, only effects metadata not data. Not wholly true and this is basically based around a misunderstanding of how BTRFS works. That is the use case it's best suited for imo. Most people can just set it and forget it (just don't forget to take advantage of snapshots and compression!). So have at it if you're looking to use it as a desktop filesystem. Snapshots are the primary biggest pro it offers, I wouldn't consider any other for Arch use.Īll said, as long as you know the limits of it, it's great, especially since it's part of the mainline kernel so no need to worry about out of tree drivers like ZFS. If you need quotas, it's slow af, so don't use those. Other than that, have at it, it's a great filesystem and I use it for just about everything: My NAS, VPS and my PC. All the other raid levels are stable if you maintain it. Believe the mkfs tools when it warns you. RAID5/6 is totally unstable, there's a plethora of bugs that makes it a non-option, don't believe anyone who tells you otherwise, even with the latest kernels. Using Btrfs as RAID on your root filesystem is also not the greatest if uptime is the goal, as you need to deal with a number of downsides both in btrfs itself and systemd (ie you need to provide the degraded mount option, which really complicates the boot process, and systemd can be problematic in that case if it's not also added to the fstab). The way to detect a degraded RAID is to monitor both the device stats for corruption and scrub.Īlso, if you had NOCOW files, scrub wont repair those, balance or cp -reflink=never is the only option way to resync them (and since there is no checksums it can never repair them). If a disk drops from the system and returns, you gotta unmount and rescan first for it to be re-added, and then trigger a scrub to repair itself. The biggest downsides it has compared to ZFS is while it is a self healing filesystem, it doesn't automatically resilver. You can add and remove disks as you please, and shrink the filesystem. Reflink copies are also pretty awesome (though if you only want this, XFS also offers it). The biggest pros it offers over ZFS are the very flexible RAID options with the ability to effectively use the total capacity of all disks thanks to its support of "degenerate stripes" and it's unique spin on raid1. The main pros are those of ZFS, it has snapshots, compression, and block level backup to remote btrfs filesystems or image files. If you don't (and honestly, if it's for desktop use, you very likely don't), then it's a great desktop filesystem and imo should be the default of most desktop focused systems. This isn't an issue for most desktop setups, but it does happen from time, especially with known buggy hardware (most USB to SATA controllers in my experience) and a few specific WD HDD firmware bugs.ītrfs doesn't have as many performance features that ZFS has, and the ones it does have are bad to use with RAID (ie NOCOW), however there are some if you need it. IF you use some form of stable btrfs redundancy, like RAID1, the other copy will repair itself. When things do go wrong, it does offer recovery options, arguably more in it's default tools than ZFS does, but like ZFS, if things go very corrupt your best course of action is mkfs and restore a backup (or btrfs restore just don't depend on that saving you, you do have backups, right?). It absolutely depends on your underlying hardware to respect write barriers, otherwise you'll get corruption on that device since it depends on the copy on write mechanism to maintain atomicity. Here's basically all you need to know when choosing Btrfs.ītrfs is a great filesystem but also greatly misunderstood.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |