[pve-devel] qemu drive-mirror with zeroinit and ceph/rbd, not working

[Alexandre DERUMIER]

>>I'll do test from a local raw 

Ok, it's working fine with a local raw.

So it's a nfs limitation


I wonder if we could add a qga "guest-fstrim" call at the end of drive mirror, if qemu disk have discard option enable ?



----- Mail original -----
De: "aderumier" <aderumier at odiso.com>
À: "Wolfgang Bumiller" <w.bumiller at proxmox.com>
Cc: "pve-devel" <pve-devel at pve.proxmox.com>
Envoyé: Vendredi 17 Juin 2016 10:55:14
Objet: Re: qemu drive-mirror with zeroinit and ceph/rbd, not working

>>the underlying file system supports SEEK_DATA/SEEK_HOLE then it should work. 

mmm, this was from a nfs server. 

I'll do test from a local raw 

----- Mail original ----- 
De: "Wolfgang Bumiller" <w.bumiller at proxmox.com> 
À: "aderumier" <aderumier at odiso.com>, "pve-devel" <pve-devel at pve.proxmox.com> 
Envoyé: Vendredi 17 Juin 2016 10:22:08 
Objet: Re: qemu drive-mirror with zeroinit and ceph/rbd, not working 

> On June 17, 2016 at 10:01 AM Alexandre DERUMIER <aderumier at odiso.com> wrote: 
> 
> 
> Hi, 
> 
> I have tested to drive-mirror an .raw file to ceph/rbd, with move disk feature, 
> 
> and it seem that the target rbd volume is full allocated. 
> 
> I'm not sure how is working the zeroinit patch ? 

Currently it relies on qemu's use of bdrv_write_zeroes() and just skips that 
while passing all other write requests through to the underlying block driver. 
(It also tracks the target file's extents to deal with out-of-order writes, 
such as zeroing a block which has previously been filled with data.) 
It should be possible to change it to check the contents of all the regular 
writes, too, but currently it doesn't do that. But if the .raw file isn't 
fully allocated and the underlying file system supports SEEK_DATA/SEEK_HOLE 
then it should work. 

Added a reminder bugzilla entry: https://bugzilla.proxmox.com/show_bug.cgi?id=1032 

> I'm seeing that it's enable in rbdplugin. 
> 
> 
> Alexandre 
>