On Fri, Jan 3, 2020 at 9:06 AM Robbie Harwood <rharwood(a)redhat.com> wrote:
Nico Kadel-Garcia <nkadel(a)gmail.com> writes:
> On Thu, Jan 2, 2020 at 2:48 PM Robbie Harwood <rharwood(a)redhat.com> wrote:
>> "John M. Harris Jr" <johnmh(a)splentity.com> writes:
>>> On Friday, December 20, 2019 10:59:52 AM MST Chris Murphy wrote:
>>>
>>>> Issuing the command once per week harms no one
>>>
>>> Based on what's actual in the Change proposal, this is not the case.
>>>
>>> Even if this goes through, in my opinion, it should only affect the
>>> GNOME Spin, or perhaps even "all graphical" spins at most.
>>
>> No? This is extremely useful for cloud environments - maybe the most
>> useful. It allows VM hosts to reclaim and reuse empty disk space;
>> otherwise, the disk images just bloat to their maximum allowed size.
>
> Its most useful for the cloud *providers*, not the cloud clients. For
> the clients, getting the AWS space pre-allocated form EBS is often a
> notable performance improvement, and restoring it to AWS saves AWS
> resources. Not the client system performance.
Sure, but in many cases the client is also the provider. Consider
running kvm on a laptop (which I and many others do for work...) - you'd
really like the disk space back you're not using, rather than each VM
taking 10-20G it doesn't need. I end up having to edit every VM
configuration in two places after each install/provision in order to get
that behavior - that's not reasonable.
By default, GNOME Boxes and virt-manager VM's, do not enable block
device discards. So this feature is a 'no op' in that case. And also
fstrim.service won't run in a container. I've updated the feature to
reflect these two things.
I'm not certain what the default should be for VMs. I really only ever
use libvirt managed qemu-kvm these days, so I'm also not sure what the
defaults are elsewhere for comparison.
--
Chris Murphy