Introducing Lima to Trellis for Faster Local Development

[swalkinshaw]

We’ve added experimental support for Lima to Trellis, which allows us to drop VirtualBox/Parallels and Vagrant for faster local dev environments.

Read more on our blog: Introducing Lima to Trellis for Faster Local Development | Roots

[MWDelaney]

Testing this right now! If it’s functional and fast, this will easily replace Parallels for my whole team!

[swalkinshaw]

We don’t have documentation for this feature yet, but the PR description has all the details including setting up sudoers and how to change the Ubuntu version if necessary: Add VM integration by swalkinshaw · Pull Request #346 · roots/trellis-cli · GitHub

[huub]

Hi, how can I connect to the database using Sequel Ace? Should I use SSH or socket?

[swalkinshaw]

Oh good question. You should use SSH. Here’s what my connection screen looks like:

1. Yyou can find SSH host and port by running this command: limactl ls (but the host is always 127.0.0.1 so you just need to find the port)
2. the username is always your local machine’s macOS username.

I’ll have to see how we can improve the trellis db open command to better work with this.

[nikitasol]

Basic usage is very straight forward and works without any problem. I have tried moving my last work to the Lima VM and after manually updating (maybe not all were required) development hosts, /etc/hosts (deleting old ones) and wp-cli alias for development (I have used limactl show-ssh --format=cmd INSTANCE to check values, but then I saw that these can also be consulted in the lima inventory file) I have managed to pull down and sync production db.

I am not sure, but I think it would be great to have a couple of additional vm commands to help with initial setup and move to the new virtual machine? Maybe trellis vm init might be good so yml file can be reviewed and updated before initial provision? And maybe trellis vm aliases for rewriting wp-cli.yml with the new values?

[nikitasol]

Also, thanks for this move to Lima. It’s absolutely much faster and so much less frustrating than with the VirtualBox.

[swalkinshaw]

Thanks for testing it so quickly!

@nikitasol

I have tried moving my last work to the Lima VM and after manually updating (maybe not all were required) development hosts, /etc/hosts (deleting old ones) and wp-cli alias for development (I have used limactl show-ssh --format=cmd INSTANCE to check values, but then I saw that these can also be consulted in the lima inventory file) I have managed to pull down and sync production db.

Removing conflicting entries in /etc/hosts and .ssh/config might be necessary (though vagrant down might automatically remove those as well).

Did you mean Ansible hosts files like hosts/development? Lima generates its own inventory file as you saw so there’s no need to remove the existing ones.

For WP-CLI aliases, I’d recommend running trellis alias which will generate them all for you. Not sure if you had used that before or just manually defined them?

[nikitasol]

@swalkinshaw

Did you mean Ansible hosts files like hosts/development? Lima generates its own inventory file as you saw so there’s no need to remove the existing ones.

For WP-CLI aliases, I’d recommend running trellis alias which will generate them all for you. Not sure if you had used that before or just manually defined them?

Yes, I thought so, but I changed hosts/development ones manually after I tried running trellis alias and got into error:

[✘] Error generating config
Error creating WP-CLI aliases. Temporary playbook failed to execute:
[WARNING]: Platform darwin on host 192.168.56.5 is using the discovered Python
interpreter at /usr/local/bin/python3.10, but future installation of another
Python interpreter could change the meaning of that path. See
https://docs.ansible.com/ansible-
core/2.13/reference_appendices/interpreter_discovery.html for more information.

The host in error was the one that was the original host ip, so I decided to change it manually to the new value. It had no effect on the trellis alias and it continued throwing the error (though showing the updated ip), so I changed the alias manually. The aliases were generated at first through trellis-cli.

[MWDelaney]

This is working great for local development, but my workflow struggles when I try to use Trellis aliases and ssh commands that work with Parallels and Virtualbox. Things like the Sync Script.

It seems like a some kind of flag containing the “current” or “last-brought-up” development environment (Vagrant or Lima), and a conditional around this line (and whatever the Go equivalent of whoami is), might get things working together?

[huub]

@swalkinshaw

limactl ls

Thanks, that’s works!

[jokanane]

Thanks, this looks promising!

Should .trellis/lima be in trellis/.gitignore? The generated files don’t look something to be committed.

[adleviton]

I’ve tested with one site. So far, so good. Note that I use neither trellis new or trellis init in case that makes a difference. I also changed the LIMA_HOME env variable so I’m working from an external drive. My site is running and was able to connect with Sequel Ace.

Questions/Comments for now:

• How can we trust local certificates (i.e. what is the new command for vagrant trellis-cert trust)?
• I tried changing settings in $LIMA_HOME/.lima/_config/default.yml (changing memory, cpus, disk size), but it seems to have no effect, even after deleting and reprovisioning. I can’t figure this out.
• A couple times, trellis vm stop didn’t completely exit back to command line.
• Not a big deal, but when I run trellis vm shell, the following errors appear:

bash: line 1: cd: /Volumes/External Drive/My Project/trellis: No such file or directory
bash: line 1: cd: /Users/adam: No such file or directory

[nikitasol]

I actually haven’t had any issue with Sync Script after manually adjusting wp-cli.yml to correctly establish ssh connection.

[nikitasol]

What I have noticed is that the port inside Inventroty (ansible_port) is changed each time I start the machine.

And as @adleviton has mentioned, for me most of times if not all trellis vm stop ends on:

INFO[0001] [hostagent] 	/Users/brew/Library/Caches/Homebrew/go_mod_cache/pkg/mod/github.com/!code-!hex/vz/v3@v3.0.4/virtualization.go:160 +0xec
FATA[0180] did not receive an event with the "exiting" status
Error stopping VM

[swalkinshaw]

Yes it should be ignored. I need to figure out a better strategy for the .trellis directory since it’s not ideal right now. .trellis/cli.yml should be committed, but nothing else in there should be. I’ll likely gitignore .trellis, and offer a new root-level cli.yml config. Feedback welcome on that.

[swalkinshaw]

@adleviton

• How can we trust local certificates (i.e. what is the new command for vagrant trellis-cert trust)?

There’s no automated way to do that yet. I might wait for SSL certificates refactor by swalkinshaw · Pull Request #1310 · roots/trellis · GitHub with https://github.com/roots/trellis-cli/pull/311

@adleviton

• I tried changing settings in $LIMA_HOME/.lima/_config/default.yml (changing memory, cpus, disk size), but it seems to have no effect, even after deleting and reprovisioning. I can’t figure this out.

The generated Lima config is in .trellis/lima/<name>.yml but there’s two issues:

1. it will be re-generated each time you re-create a Lima instance
2. I don’t even know how or if Lima lets you change those params on an existing instance with re-creating it

Basically, I need to add support to the CLI config file so customizing those VM properties so they’ll be in the generated config. Right now only the defaults are supported as you found. If you want to try contributing that change, it shouldn’t be so hard and I could help out.

@adleviton

• A couple times, trellis vm stop didn’t completely exit back to command line.

I’d recommend posting an issue on https://github.com/lima-vm/lima/issues since that’s coming directly from the vz package that Lima uses for macOS virtualization.

Reminder: trellis vm mainly wraps underlying limactl commands so you can always run those manually too.

@adleviton

• Not a big deal, but when I run trellis vm shell, the following errors appear:

Yeah I get that too. Lima assumes those directories are mounted by default, but they aren’t in my integration. So there’s two options:

1. keep those mounted default dirs to satisfy Lima’s assumptions
2. customize the default shell dir which I’ll look into. I think a good default is to auto cd into /srv/www/<main site>

[swalkinshaw]

@nikitasol

What I have noticed is that the port inside Inventroty (ansible_port) is changed each time I start the machine

I’m using Lima’s default method of auto finding a free SSH forwarding port. It’s good because you don’t have to worry about manually setting it or having conflicts with multiple machines, but it does mean the port isn’t stable.

For things that trellis-cli can control (like trellis db open, which I’ll fix soon), it will automatically use the generated inventory file ensuring everything works. But for external tools like sync script, it does mean you need to update the port manually each time

Maybe I could make an SSH port option if people want to hard code it?

[adleviton]

@swalkinshaw

The generated Lima config is in .trellis/lima/<name>.yml but there’s two issues:

1. it will be re-generated each time you re-create a Lima instance
2. I don’t even know how or if Lima lets you change those params on an existing instance with re-creating it

Basically, I need to add support to the CLI config file so customizing those VM properties so they’ll be in the generated config. Right now only the defaults are supported as you found. If you want to try contributing that change, it shouldn’t be so hard and I could help out.

I’d like to help but not sure how or where to get started working on a CLI package. If there is a crash-course out there, let me know. For now, my confusion here comes from trying to follow their instructions for default.yml. Specifically this part:

# The builtin defaults can be changed globally by creating a $LIMA_HOME/_config/default.yaml
# file. It will be used by ALL instances under the same $LIMA_HOME, and it
# will be applied on each `limactl start`, so can affect instance restarts.

[swalkinshaw]

Oh… I’d actually think that should take affect since it’s defaults. I’ll look into that first.