ZFS vdev expansion or Snapraid option with other filesystems

[Orbixx]

This is a feature request that naturally appeals to me, otherwise I wouldn't be writing it - but I assume there are a lot of NAS users out there like me who want to gradually expand their array as and when the storage requirements come up, predominantly because of money and the fact that they (I) don't know exactly how much I'll end up actually using.

I hear that Nexenta have somehow implemented vdev expansion, which would be enough to satisfy the above requirements, but sod knows when they'll contribute it back to illumos - maybe they provide the source for their ZFS modifications on their website.

Failing that, Snapraid with some other common filesystem support would also fit the bill I believe, and provide a bit more flexibility especially for those that want to bring their existing JBODS to FreeNAS.

I have no evidence to back up what I'm about to say at all, but I reckon if you provide support for this kind of user you'll see a lot more use of FreeNAS. Whether the above agrees with your vision of where you want FreeNAS to go or not, is entirely something else however and I would understand a refusal to Snapraid and other filesystems to go with it.

Vdev expansion would be super nice though. Also, first post, hello!

Edit: It seems my comment about Nexenta and vdev expansion might just be hearsay, but here's an interesting post from Adam Leventhal 2 years before he left Oracle surrounding the topic and how it can be done: https://blogs.oracle.com/ahl/entry/expand_o_matic_raid_z

 

[cyberjock]

That's a feature that's been discussed until the horse was dead. There's no simple way to implement it(along with many ZFS features), there's very few people on the planet with the knowledge to even implement new features and even fewer that will do it for free on their own time for an open source community. One of the great things about Sun was that many of their projects were open source despite making a profit off of it. ZFS was designed to be an enterprise class file system. It is exactly that. Enterprises NEVER add single disks. They add large bunches of disks at the same time, which is exactly what a vdev is.

Also, Snapraid and some of the others may "fit the bill", but they will never ever have the reliability of ZFS. So your comparison is apples-to-oranges. ;)

In short, you can expect the feature somewhere around 2 weeks after the zombie apocalypse. Lots of people would love to see the feature, but its just not very feasible and not something FreeNAS developers could ever come up with. It's the ZFS experts that would have to do it.

Welcome to the forums!

 

[Orbixx]

I figured you might say that. It's a bit frustrating, given that any other NAS solution that comes close seems half-baked and the only thing I can criticise about FreeNAS is the inability to expand an array freely.

 

[cyberjock]

I figured you might say that. It's a bit frustrating, given that any other NAS solution that comes close seems half-baked and the only thing I can criticise about FreeNAS is the inability to expand an array freely.

/grin

They are 1/2 backed. There's a reason why a full blown Windows NAS from a major manufacturer costs thousands of dollars, the 1/2 backed NAS solutions are cheaper(they skimped on doing things right), and FreeNAS(while the OS itself is free) has needs that must be fulfilled to protect your data safely and securely for the long run.

Those 1/2 baked systems are bent on convincing you to buy them. When your data is lost they won't coddle you. They want the sale, and if you're convinced that those really are trustworthy, you are in the monitority(and probably are more concerned with saving a buck than saving your data). Remember, they have you they second you take their product home. FreeNAS on the other hand puts priority in protecting your data with ZFS.

 

[Orbixx]

Well, I suppose one can easily test the capability of their software recovering data through realworld testing of simply pulling disks out whilst IO is happening. The lack of a non-terrible all-round good solution to this problem is making me seriously contemplate writing my own filesystem, or some kind of raid that works above the FS.

Thanks for responding to me anyway mate, it was good to get some realistic insight.

 

[cyberjock]

And apparently auto-correct changed my baked to backed a few times. Wonder what my typo was.

Building a FS is no easy feat. There's so many gotchas, edge cases, etc. I've heard there's less than 10k people on the planet with the knowledge and skill to even try to build a file system. ZFS was a team of like 40 people and many years of development if what I read and remember is true.

Personally, despite the "limitations" of ZFS, I find that its pretty amazing. There is potential for BTRFS to be a very good option in a few years. It really just depends on how well the developers start adding in more features and such. BTRFS is also linux only for the moment, so there's quite a few hurdles before we could possibly see BTRFS as part of FreeNAS.

 

[paleoN]

Edit: It seems my comment about Nexenta and vdev expansion might just be hearsay, but here's an interesting post from Adam Leventhal 2 years before he left Oracle surrounding the topic and how it can be done: https://blogs.oracle.com/ahl/entry/expand_o_matic_raid_z

Ah yes, the mythical Block Pointer Rewrite. They were "working on it" for years with no news. Finally, one of the core ZFS devs mentioned that BP rewrite would be complex to implement, touches a lot of the code base and had no time or much interest in pursuing it.

The Nexenta devs did recently talk about it and would be one of the groups actually capable of implementing it. Perhaps, we should all start holding our breaths. Oh, wait.