On Thu, 2019-11-14 at 22:03 -0500, Steve Litt wrote:
> The piece of information I couldn't find in your (John Morris') data
> is
> how much of the 160GiB is consumed with data. It makes a big
> difference.

They have had most of their space written to at this point. The setup
is three physical machines that host two virtual machines (web frontend
and a postgresql database) running by themselves on two of them and
another pair of development VMs on the third or the third is powered
down as a spare. The VMs rotate between the three physical hosts on
every upgrade cycle to spread the wear on drives and the servers. The
production images are about 32GB, between 50% and 80% utilized
currently, but the remainder of the drives are unused, but dirty,
between moves. Can't use fstrim in the VMs because qemu / libvirt
doesn't support it yet.