I can't figure out how disk allocation inside the ...
# rancher-desktop
e
I can't figure out how disk allocation inside the container is managed with RD. Docker Desktop, Lima, Colima have explicit settings for it. Orbstack seems to let it float somehow. Here it seems to be about 100G
Copy code
$ ddev exec df -h /
Filesystem      Size  Used Avail Use% Mounted on
overlay          98G  7.8G   86G   9% /
f
I don't understand the question
Are you asking about setting the disk size for the VM? That is sadly still missing. There are manual ways to resize (well, grow only, not shrink) the disk using qemu-img, if that is what you are looking for
See the thread right here from yesterday, and how changing the disk size throws an error in the latest release (plus a workaround)
I get that this is totally unsatisfying
e
When people use DDEV for a while without cleaning up images or volumes, etc, have huge databases (volumes) etc, they end up getting our warning. With the other providers, we can tell them what to do about it.
So it's just fixed, unless you fiddle with Lima inside?
f
You can run this command and restart Rancher Desktop:
Copy code
echo "disk: 500GB" >"$HOME/Library/Application Support/rancher-desktop/lima/_config/override.yaml"
Just make sure you never decrease the size. Once you restart Rancher Desktop the disk size will grow.
πŸ‘ 1
But
docker system prune
is still recommended to recover space
e
I know how that's documented in Lima, does Rancher have a doc for it?
f
No it is not, this really should be exposed in the preferences, but we never get around to it. Maybe we should add this at least to the FAQ
e
f
I'll create a docs issue...
e
Until we added the warning a few years ago people would always get mysterious disastrous un-troubleshoot-able results at the end of their disk. Since then we don't, but we don't give them enough guidance
f
Yeah, once you run out of space, it is much harder to recover because
docker system prune
may be unable to run. You need to have at least some space left. The fix then is to increase the volume size at least a little, and then prune images right away before you run out again
e
And of course, thank you for keeping Rancher Desktop so stable over the years, even when things might go wrong in its components. So glad it doesn't automatically follow lima right off the cliff when something happens there.
f
Yeah, we are a bit conservative in updating things because having to do patch releases is too much work πŸ˜„ The whole waffling between SSH and GRPC forwarding is a bit embarrassing. But some bugs you only find once many people use the tool.
πŸ˜€ 1
So we almost enabled GRPC forwarding in the 1.20.0 release, but got a bad feeling about it and pulled it at the last minute. You can still set it manually from the CLI for testing purposes.
e
Glad you held back πŸ™‚
f
Yeah, but I hope it will mature soonish. It is blocking UDP support (and inotify)
πŸ‘ 1
e
inotify works fine with mutagen πŸ™‚
I guess it's time for me to do another performance comparison of all the docker providers.
f
I think OrbStack will come out on top. Unfortunately it is all closed source.
e
I see I can lose some of the permutations next time, just do virtiofs for everybody. Haven't tried the other DD alternative
w
may want to include container framework on 26 to see how that keeps up
e
Do you have any idea how to test that @wide-mechanic-33041? They don't have any real tooling for it do they?
w
if you are running 26 than https://github.com/apple/container
e
I am running 26, but they don't have any docker wrappers or compose or anything right?
w
f
For some reason they are reinventing the whole user space
w
also known as being apple.
i expect there will be shims written as well
e
An alias doesn't work when DDEV is set up to use docker API 😞
As is the rest of the world
f
Unfortunately true. Which is also blocking the way to moving to
containerd
for many people
πŸ‘ 1
e
Thanks!
w
and don't hold your breath on containers, but will see how open it really is. lots of folks are saying the same thing though around api as critical path. just like i am hoping w wsl going open that wsl gets a beefier api as well so its not always shell parsing