Comment 31 for bug 1646181

I am afraid this issue was prematurely marked fixed. As far as I can see it still exists in at least Ocata, just with a shifted threshold. I have a snapshot image that is 400GB in virtual size, 124GB qcow2 file size, on NFS, and I am seeing the prlimits error even with --cpu=8.

http://paste.openstack.org/show/719357/

Between lines 10 and 11, it takes 25 minutes to transfer the image to the compute node. Once the transfer is complete, qemu-img info fails and we never get to spawn the instance.