I see that BTRFS is in 7.2 and still listed as a technology preview. Is there an estimated VE version when this will be stable for production use?
Hopefully, when it will be production ready .. if that date ever comes ... current status is here.
From all the filesystems around, those two are very similar in all respects that count and BTRFS was started as a clone of ZFS back in the days when ZFS was not open sourced. It is also a Copy-on-Write filesystem with checksums and stuff, so yes, it'll also wear our cheap consumer SSDs. The problem with wearing out is a purely PVE-specific problem due to the /etc/pve filesystem and the RRD graphs. If you disable both (or put on another disk), you will not wear it out so much.
Or you could use it for testing and troubleshooting purpose.
Until now I used qcow on dir storage (on zfs) and lvmthin for this purpose, mainly because you can use snapshot hierarchy with both of them.
But since I've never had any painful issues running btrfs with openSUSE Tumbleweed (btrfs for /) for several years on my laptop, I just decided to use btrfs storage on pve and see if that works out.
Pleasant part is: btrfs storage on pve supports snapshot hierarchy
Sorry, i am very new to proxmox and i was looking around to read about the wear out of btrfs drives in combination with proxmox, thats why i came to that thread.
What you are saying is not correct. BTRFS has nothing to do with ZFS and they are only similar by looking at the features but not at how it is done.
If BTRFS would have started as a ZFS clone, it would be at least a 128bit FS as ZFS is.
BTRFS is based on a binary tree structure and has absolutley nothing to do with ZFS.
Based on that, i am not sure to trust any of your statements actually - i will need to do some more proxmox research then.
Sorry, I was not clear enough. You're totally right, it's a "clone of the features" and of course it's not a carbon copy clone. So the only word missing is "as a FEATURE clone of ZFS". Everything else is still true.
No, it's a B tree, not a binary tree. See the original paper, in which the actual similarities and dissimilarities to ZFS are also shown. Therefore my "clone of ZFS" statement.
Because the developers have not stated otherwise. Important things like RAID5 is still not production ready. I don't want to use a filesystem that states "mostly OK" for features available in other filesystems.
They are being extremely conservative saying that. Like anything else you need to test it for your use case. Lots of people using it in production. I don't think Facebook would rely on Btrfs for their infrastructure if they didn't think it was ready.
Last edited:
I even highlighted it in the other post:
As I said above, before you edited your post, simply use it and do not care about the status at all; but do not complain, if something goes wrong...
Why are you contradicting yourself? First give reasons not to use it then saying just ignore that?
It's not integrated in the GUI yet. That, and calling it a technology preview, is a Proxmox specfic problem that is not solved.
Last edited:
I only quoted two statements from a Proxmox developer in my first post. Nothing more, nothing less.
Before you edited your post, your answer to this was:
and even after your edit, your answer is:
So you completely ignored the part, that I highlighted on purpose:
Since you only touched this single raid5/6 topic and tried to negate it, ignoring all other topics, it sounded to me, that the only concern you have, are the two words: "technology preview".
That is, why I played along your somewhat ignorant "Problem solved" game...
Oh, honestly?!
One of the points, why it is still a technology preview is, that the integration into PVE is not completed yet, as I now quoted three times! Is it so hard to understand?
Sorry, but I am tired of this discussion.
Either use it as it is right now or not. Its current status is what it is anyway.