[pve-devel] pve-zsync and mount points

[Fabian Grünbichler]

On Sat, Mar 25, 2017 at 05:31:09PM +0100, Stephan Leemburg wrote:
> Hello All,
> 
> I am new to this list. My name is Stephan.
> 
> I am setting up pve-zsync and ran into some issues.
> 
> Before I start sending patches, I want to check with you all.
> 
> I have a container that has:
> 
> mp0: /tank/vmdata/subvol-homes,mp=/home
> mp1: /tank/vmdata/subvol-mail,mp=/var/mail
> 
> With the current pve-zsync this does not work.

yes, as those are bind mounts where we don't know that ZFS is the
underlying FS.

> 
> I already patched a little bit, but it is not precise enough.
> 
> What I would like, and I can supply the patch if we agree, is:
> 
> 1) have pve-zsync accept (with a flag or so) to ignore mp\d+ entries

this should AFAIK be possible by setting "backup=no" on those mount
points.

> 
> 2) if not ignored, then do something with it, and possibly in- and
> exclude some types
> 
> For 2) I think that the syntax for the .conf files should be extended.
> If we could use somethings like:
> 
> mp0: nfs:/path/to/share,mp=/target/path
> 
> mp1: vmdata:subvolume-name,mp=/target/path
> 
> mp2: fstab:/path/to/mountpoint,mp=/target/path
> 
> etc.

you mean, use PVE's storage system[1]? it does exactly this - first
comes the ID of a defined storage (which means we also know which type
of storage), then the ID of the volume itself. E.g.[2],:

mp0: myzfsstorage:subvol-vm-100-disk-1,mp=/path/in/container

> 
> Then things would become easier for zsync. And it would be more generic
> and consistent with fsroot.
> 
> Looking forward to your comments.

see above, hope it helps to clear up some potential misunderstandings ;)

1: http://pve.proxmox.com/pve-docs/pve-admin-guide.html#chapter_storage
2: http://pve.proxmox.com/pve-docs/pve-admin-guide.html#pct_mount_points