[pve-devel] cache unsafe write for qemu-img convert, vma extract and vmstate snapshot

[Alexandre DERUMIER]

for the vmstate, on a ceph ssd cluster, I have now 10s vs 50s for 32gb vmstate.



----- Mail original -----
De: "aderumier" <aderumier at odiso.com>
À: "pve-devel" <pve-devel at pve.proxmox.com>
Envoyé: Lundi 1 Août 2016 08:50:59
Objet: [pve-devel] cache unsafe write for qemu-img convert,	vma extract and vmstate snapshot

As reported by Eneko Lacunza, 
ceph/rbd by default use writethrough even if writeback is defined, until a flush is detected. 

We currently use writeback without sendind any flush for 3 things: 

qemu-img convert : (default is unsafe, but we force writeback in qemuserver) 
vma restore on target destination 
vmstate snapshot 

The patches serie set unsafe opening (BDRV_O_NO_FLUSH). 

And also fix the rbd drive to disable cache_writethtrough_until_flush if cache=unsafe 

@Eneko Lacunza : Can you test it ? I have rewrote your last patch (http://pve.proxmox.com/pipermail/pve-devel/2016-July/022226.html), 
to handle vma extract and vmstate snapshot 

_______________________________________________ 
pve-devel mailing list 
pve-devel at pve.proxmox.com 
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel