I will, reluctantly, try to answer - This is how ZFS or systemd flame wars starts - it brings the worst of the people.
To be honest, I am truly alarmed by you bringing zfs and FUD into this - it seems that one hand you are asking because of no experience with btrfs and ZFS - then in the same breath you already know that it is not good enough and too fancy ..... You get the point of what I am worried about. 1. I only have good experience with btrfs, I use(d) it personally and at scale in enterprise. 2. My needs are pretty basic as far as btrfs goes - basic native redundancy, like most - I use mdraid for raid, deduplication, do not use compression, CoW, snapshots, snapshots, ..... did I say snapshots enough, volume resizing both up and down, adding and taking away storage, power cuts, forced power downs, sub- volumes, not CoW for VM block storage, SSD TRIM. 3. OS disks: It is default openSuSE/SLE install choice for about 5 years now - it is huge added value due to well integrated snapshots and roll-back. In short - every update, install or yast config change creates new snapshot - you can totally boot into then by picking one from Grub menu. You can restore snapshots on running system in seconds without the hassle of restoring from a backup disk. When experimenting with SW install and/or OS config - I can install, configure things, experiment as much as I like then revert to an earlier shapshot returning the OS to the original stable state. 4. Bulk Data storage: .snapshots take fraction of a second to create and destroy. They are essentially free. You can make a snapshot every hour, keep 1/6 of them per day, then 2 a week, one per month ..... it does not slow down anything, it does not duplicate data. All you need is the disk space for deleted /changed data until it falls off the snapshot train. 5. Backup: You can send snapshots to remote btrfs storage - this is totally awesome - make a snapshot send it to backup filer/target - only the differences from previous snapshot gets transferred. You change 10GB in 10TB volume - only the 10GB is transferred - no need to crawl the file systems on both ends with rsync + compare and then transmit the whole files which changed. btrfs snapshots knows what bytes changed - it only sends them and fast. 6. raid - use mdraid on linux like everyone else - then create volume on the top of that and format it with btrfs. Raid is not the hot selling feature for both btrfs nor zfs - redundancy is - you just tell the file system how many data replicas you want - btrfs/zfs will oblige. 7. no CoW for VM block storage - does not even need own sub-volume anymore - it is just an attribute on file or directory. The same applies for native compression. 8. You can add and !remove! (shrink) storage from volume that is awesome. 9. reliability - I run weekly scrubs - have not experience any error reports, ever, in my 5-ish years using it. And it has lost power, and it has lost disks. In the same time, I have experienced maybe 2-3 ext3/4 fsck fixes - which probably means that, in my experience, it is as reliable as ext4 or reiserfs I used to use in pre-mature-ext journaling days. Most of the time I do not know about btrfs - until I accidentally delete/overwrite data or update/install something I would rather not. Only then - I know about btrfs. I do not think about ransomware - having configured regular automatic data snapshots - I guess, I would run out of diskspace if something tries to encrypt all my data - the old data would be in the snapshots for weeks - scrambled data would fill the disk(s) alerting me to the fact. Most of the above - maybe with the exception of shrinking volumes - applies to also to ZFS. I do not run solaris or *bsd* anymore and the big O corp will probably remain legal threat forever (just ask anyone running their SQL or the big G corp how they love it) - so it makes sense to use btrfs on linux. If Fedora chips in - it might get even better at faster rate. I am not sure if this non technical opinion piece is what you asked about - btrfs is just a modern file system - install openSuSE from scratch and you will be using it too. If you do not know why you need it - you will not notice it - unless something breaks you OS install or you delete a file accidentally. Then, you will probably realize that you have plenty automatic OS snapshots, but you did not configure any snapshots for your data. Hope that helps, Tomas On Tue, 2020-08-11 at 21:12 -0700, Tom wrote: > Would you mind sharing your experience if you go this route? I tried > btrfs a few years back in my testing did not find it to be mature or > reliable enough to trust my data with and continued using OpenZFS. I > don't know if much has changed but I think the raid write-hole is still > a problem for btrfs. If I were you and your not doing anything fancy > where ext4 works for you I'd keep that. If you require fancy filesystem > features I'd use ZFS instead of btrfs. But please do let us know how > btrfs works for you. > _______________________________________________ PLUG: https://pdxlinux.org PLUG mailing list PLUG@pdxlinux.org http://lists.pdxlinux.org/mailman/listinfo/plug
