I'd like to create a copy-on-write qcow2 image on top of a read-only base image. However, the write performance of the copy-on-write top image is too slow for the intended workload. Is there a way to preallocate a portion of the copy-on-write image at creation time? I don't want to preallocate the whole thing, but a few tens of GBs should ensure we remain scalable without taking a performance hit the majority of the time.
feedback
|