[bitfolk] Re: Sending emails from Bitfolk VPS

Hello, On Fri, Oct 14, 2022 at 04:39:42PM +0100, Andy Bennett via BitFolk Users wrote: Is the RAID-5 write hole not solvable in software? For a long time, Linux MD parity RAIDs (so -4, -5, -6) have defaulted to using a write-through cache which is supposed to solve the write hole: https://www.kernel.org/doc/Documentation/md/raid5-cache.txt https://lwn.net/Articles/665299/ I thought that non-volatile cache on RAID controllers was so that you could safely accelerate writes as otherwise you have to disable the write cache that is in the HDDs themselves as they won't survive a power loss (and is also how the software in hardware RAID controllers closes the parity write hole). If not using traditional RAID (hardware or software), it depends. ZFS solved the parity write hole by using copy-on-write: https://pthree.org/2012/12/05/zfs-administration-part-ii-raidz/ notably btrfs has not yet solved it for the parity-based redundancy levels, and that is one of the reasons why parity-based redundancy in btrfs is not recommended even as recently as last week: https://lore.kernel.org/linux-btrfs/86f8b839-da7f-aa19-d824-06926db13675@gm… But anyway, I don't think people using MD Linux RAID-5 on a modern kernel with HDD write caches disabled need to worry about about write hole, unless they specifically went out of their way to disable the MD cache. HDD write caches default to "on", by the way… Cheers, Andy -- https://bitfolk.com/ -- No-nonsense VPS hosting