Am Wednesday 08 December 2010 schrieb Brian Jackson: > qcow2 (I hope you are using that vs just qcow) is known to be a tad on the > slow side on metadata heavy operations (i.e. mkfs, installing lots of > files, etc.). One trick some of us use is to use the -drive syntax (vs > -hda) and set the cache option to unsafe or writeback for the install > process. The other alternative is to use preallocated raw images (i.e. > made with dd vs qemu-img). I've been informed that in 0.12.5 the writeback > trick won't do any good due to some extra fsync()s. So your best bet is to > upgrade to 0.13 and use cache=unsafe. This helped, thanks a lot. (Maybe this could be documented at a more obvious place? I tried googling quite a lot for the solution - a faq entry explicitely mentioning mkfs? It seems a quite common task to me) -- Hanno BÃck Blog: http://www.hboeck.de/ GPG: 3DBD3B20 Jabber/Mail: hanno@xxxxxxxxx http://schokokeks.org - professional webhosting
Attachment:
signature.asc
Description: This is a digitally signed message part.
