Playing with experimental features: btrfs

[liszca]

Playing around with btrfs I noticed replication doesn't work.

Here is my questions;

• Its not working because its not fully implementet
• Its not working because I am using a mounted btrfs partition (didn't think about a possible subvolume approach if there is any)

Is this where the experimental feature stands or can I get it to work by configuring it differently?

 

[Dunuin]

Playing around with btrfs I noticed replication doesn't work.

Here is my questions;

• Its not working because its not fully implementet
• Its not working because I am using a mounted btrfs partition (didn't think about a possible subvolume approach if there is any)

Is this where the experimental feature stands or can I get it to work by configuring it differently?

Not sure if it supports replication at all. But if you mounted a btrfs partition and you are using it as a directory storage, then PVE won't be able to make any use of the btrfs features. A directory storage for example uses the snapshot capabilities of the qcow2 file, even if that qcow2 file is backed by something like btrfs or ZFS that would support native snapshotting.

 

[rason]

Replication currently is limited to ZFS. Here is the feature request on bugzilla: https://bugzilla.proxmox.com/show_bug.cgi?id=4193

 

[liszca]

Not sure if it supports replication at all. But if you mounted a btrfs partition and you are using it as a directory storage, then PVE won't be able to make any use of the btrfs features. A directory storage for example uses the snapshot capabilities of the qcow2 file, even if that qcow2 file is backed by something like btrfs or ZFS that would support native snapshotting.

right now I am playing with LXC machines. There snapshots do work, but how should I add my btrfs partition to make full use of its features? Is creating a subvolume the way to go?

 

[Dunuin]

I guess you would need to add a btrfs storage: Datacenter -> Storage -> Add -> BTRFS

 

[liszca]

I guess you would need to add a btrfs storage: Datacenter -> Storage -> Add -> BTRFS

I tried that, but I couldn't provide it the correct partition intended for use

 

[janssensm]

Did you follow the steps provided in the admin guide?
https://pve.proxmox.com/pve-docs/pve-admin-guide.html#chapter_btrfs

If you follow that, it should work.
And for the experimental: In the docs it's called technology preview, not experimental.
The basic stuff will just work, including snapshots.

 

[liszca]

Did you follow the steps provided in the admin guide?
https://pve.proxmox.com/pve-docs/pve-admin-guide.html#chapter_btrfs

If you follow that, it should work.
And for the experimental: In the docs it's called technology preview, not experimental.
The basic stuff will just work, including snapshots.

That guide helped a lot, thanks for the link.

In addition there is an obvious command which did show me it must be working:

btrfs subvolume list /btrfs1

In my case outputs:

ID 261 gen 1817 top level 5 path images/10011/vm-10011-disk-0
ID 262 gen 1817 top level 5 path images/10012/vm-10012-disk-0
ID 263 gen 1817 top level 5 path images/10013/vm-10013-disk-0
ID 264 gen 87 top level 5 path images/10011/vm-10011-disk-0@ttest

The last one is a Snapshot, looks good to me :]

Next is how PBS likes BTRFS

 

[_gabriel]

PBS bypass storage backend on source and destination, it has its own technology to store backups in many files ( in .chunks folder) on regular filesystem.
his backups snapshots/versionning/history doesn't require qcow2 / lvmthin / btrfs or zfs.
ext4 or xfs is the recommended fs for pbs datastore.

 

[Dunuin]

PBS bypass storage backend on source and destination, it has its own technology to store backups in many files ( in .chunks folder) on regular filesystem.
his backups snapshots/versionning/history doesn't require qcow2 / lvmthin / btrfs or zfs.
ext4 or xfs is the recommended fs for pbs datastore.

ZFS/btrfs also got its benefits...PBS can only detect data corruption with its verify jobs but can't do anything against it. While filesystems like ZFS below the datastore could also repair the damaged chunks. But yes, from a performance standpoint a non-CoW filesystem like xfs/ext4 would be faster.