[sheepdog-users] Failed to read, No object found

Liu Yuan namei.unix at gmail.com
Sat Jan 5 08:42:04 CET 2013


On 01/04/2013 11:18 AM, Liu Yuan wrote:
> On 01/04/2013 12:30 AM, Valerio Pachera wrote:
>> I got this error on a disk that I know it was corruped after a 
>> serie of tests. I was able anyway to covnert it to qcow2; qemu-img 
>> check was giving no errors and the guest was running fine from the 
>> qcow2.
>> 
>> In a case like this, is there a way to fix the vdi or do we have to
>> convert it out and in again?
>> 
> 
> Seems that using qemu-img conversion back and forth for now is the 
> only way to work it out to be a sub-normal working state. This means 
> the blocks unfortunately mapped onto the 'not found object' that OS 
> previously write will be lost and zero will be returned when those 
> blocks are read again.
> 
> Maybe we should add another option to 'collie vdi check' to work 
> broken VDI into sub-normal state.
> 

Ah, I remember this is a bug of QEMU. which version do you use? The fix
is merged into the upstream QEMU now but the upstream QEMU has other
problems for Sheepdog to work perfectly with object cache: that is, the
upstream QEMU including v1.2.x can only enable writeback mode for object
cache. (I'm now struggling to fix it).

Thanks,
Yuan



More information about the sheepdog-users mailing list