Return to



Yup, and you can also have a mix of different sized VDEVs in a pool and ZFS will balance across them.

For quite some time i was running with 2x mirrors of different sizes. Performance may not be optimal, but if you just need to expand space, so long as you can get enough disks to fully upgrade a single VDEV in your pool (or even add another VDEV), you can.

If all you care about is resiliency (vs. performance - e.g., for archive/backup/nas type storage), then adding a VDEV to an existing pool is no big deal.



Thats is the current mentality. Make mirrored pairs and upgrade 2 disks at a time.

If you have a set of say 7 disks in raid z2. Not untill every disk in the 7 disk pool is upgraded can you resize so to speak.

BTRFS lets you mix and match anything and even rebuild the drives as raid 0 or 1 and 5 and 6 but 5 and 6 are possibly broken.

Its like to actually finish a FS is too hard. Start a new one because its is easy to have excuses your FS it broken. Its beta.



Sure that’s a valid complaint, unlike “fixed disk sizes” which is incorrect.

1 Like


i believe i read or heard from a good source that expanding vdevs with more drives is something ZOL wanted to do.
This was less then 1.5 year ago.



I find myself happily incorrect all to often. There is much to learn.



The thing is, there are very few people who have the experience and mentality required to get this stuff right.

A lot of people think “oh, that’s an easy problem!” and then when they get into the weeds they realise that actually, no it isn’t.

IMHO, ZFS has the correct mindset and some of the “limitations” may be annoying, sure, but mostly come down to “don’t make dumb decisions when setting up your pool, by being cheap” and “reliability for 24/7 operation isn’t entirely free”.

If you want max performance just run RAID0

If you care about long term storage of your data, then ZFS is imho the only real option.

also, ZFS originated at sun, with some very clever sun engineers. hence, i trust the design is mostly right. i think the proof is in the pudding… ZFS on linux less trustworthy IMHO… but still better than EXT as the design principles are right, you’re just dealing with implementation bugs.



I agree. In a world with torrent pirates and huge fille downloads of Game of Thrones to day trading where nanoseconds matter.

Intel is bleeding the FS into ram with optane and persistant memory. What the hell FS will that be :stuck_out_tongue:

Im sure it will be a kernel ,memory management thing. Not a file system cause FS are broken.



you’ll be thrilled to learn about NVDIMMs then

1 Like


Thread title should be…

ZFS vs *

1 Like


ZFS needs one of those “calvin urinating on whatever” memes.

1 Like


and when your file is lets say a 5gb h.264 mp4? you just store a hash for every single block individually so you can report if that entire file is wrong when one block goes bad because you are the aryan file system?

bow down to the zfs master race try our punch with our zealot robes but leave you’re mongrel filesystems at the door

you dont want to end up like these bois



love that a shitpost got this much engagement and detailed good faith discussion but at the end of the day ZFS is the best at what it does, but only a small amount of people need its features or anything in its class of filesystem

If you dont need ZFS features, don’t run ZFS

If you do, it’s probably the best option.

1 Like


No when the data on disk is wrong it can be corrected using the parity or mirrored copy or just another copy (however you choose to set it up).

If you have a collection of large media files, you can create a dataset with a larger block size, too, so you don’t have as much overhead for metadata. Likewise if you have a collection of small files, you can create another dataset with a smaller block size, so you avoid wasting a bunch of empty space in each block. Very small files can even be fit into the space of a metadata block itself, avoiding the need for allocating a block for storing the file contents at all.

I feel like you’re the one singling out ZFS as an inferior race of filesystem. I’m not sitting here making baseless criticism of ext4, or hassling people who use XFS. Each filesystem has its own merits and applications.



ZFS is a storage FS but consumers are fast smashing into storage sized FS problems.

Well so many normies record 4k video on a phone which means large files by default. It will only get bigger.

So unless your network via an ISP allows the cloud to work it out then the file systems on our devices needs to catch up to huge files that are non compressible. I know a log files on a linux box are super compressible.

One family holiday could create a 1TB of content. If Dad was camera happy. Well somewhat of an exaggeration.

File systems are now more than ever needed to hold and works with more data. Keep it safe.



What do you mean by this?

ZFS has larger capacities for file size, partition limit, etc. than pretty much anything else and has features for changed block tracking built in.

What do you think will do a better job than ZFS moving forward, and why?

ZFS is used by CERN for their hadron collider research and i’m pretty sure they have bigger datasets than any of us. Probably bigger than all of us here at level1 put together.

The only real argument i’ve seen against ZFS is that it is “expensive” in terms of resource requirements (but offers integrity features that nothing else does). But as time moves on, hardware gets cheaper and those resource overheads become less relevant.



Well the “resource” ZFS requires on top of the most popular devices people have now are phones and there pushing 128G and 256G. It won’t be long till there 1TB and computers are already 1TB or more with a HD.

Files systems like EXT, FAT, NTFS were not designed to manage TB’s of storage over time. In fact they dont even detect most errors.



ZFS is designed to handle these sorts of capacities however, hence my confusion regarding you calling it a “storage fs” and that consumers are fast running into “storage sized fs problems”.

Whatever that means.



Do you mean something like ZFS is not best suited to the resources that phones/tablets have, and the way they interact with their storage?

I can see the point, and maybe F2FS might be better for a phone/tablet; Beyond that, I reckon ZFS would shine



I do recall a tweet from Adam (surname escapes me, @ahl on twitter - one of the original ZFS developers) that ZFS was originally deployed on hardware with similar spec to the original iphone…



Leathenhall maybe?
Was he the guy who spilled the beans about apple switching to ZFS before Apple did, causing them to stop the rollout, and create the abomination of AFS(or plus?)