On Mon, Mar 22, 2021 at 01:47:18PM +0100, Mischa wrote: > > > > On 22 Mar 2021, at 13:43, Stuart Henderson <s...@spacehopper.org> wrote: > > > >>> Created a fresh install qcow2 image and derived 35 new VMs from it. > >>> Then I started all the VMs in four cycles, 10 VMs per cycle and waiting > >>> 240 seconds after each cycle. > >>> Similar to the staggered start based on the amount of CPUs. > > > >> For me this is not enough info to even try to reproduce, I know little > >> of vmm or vmd and have no idea what "derive" means in this context. > > > > This is a big bit of information that was missing from the original > > Well.. could have been better described indeed. :)) > " I created 41 additional VMs based on a single qcow2 base image.” > > > report ;) qcow has a concept of a read-only base image (or 'backing > > file') which can be shared between VMs, with writes diverted to a > > separate image ('derived image'). > > > > So e.g. you can create a base image, do a simple OS install for a > > particular OS version to that base image, then you stop using that > > for a VM and just use it as a base to create derived images from. > > You then run VMs using the derived image and make whatever config > > changes. If you have a bunch of VMs using the same OS release then > > you save some disk space for the common files. > > > > Mischa did you leave a VM running which is working on the base > > image directly? That would certainly cause problems. > > I did indeed. Let me try that again without keeping the base image running. > > Mischa
I seemed to recall that the base image is not supposed to be modified, so this is a pretty big omission. Per original commit message: "A limitation of this format is that modifying the base image will corrupt the derived image." https://marc.info/?l=openbsd-cvs&m=153901633011716&w=2 -Bryan.