How to expose colima env via tcp?

[mikhail-gaponov]

Hi Colima Community!
I've tried to find out a way to expose colima profile runtime via TCP and I didn't find it at colima docs or Internet.
Could somebody show me the right direction to move forward?

As for the final goal I'd like to have separate local docker environments.

[abiosoft]

What do you mean by expose via TCP?
If you referring to a reachable IP address, then that can be enabled via the --network-address flag.

colima start --network-address

You can view the IP address for each profile with

colima list

[mikhail-gaponov]

Well, how deep the rabbit hole is.... I have both needs with next schema:
I have 1 Macbook and from 1 to 3 linux-based bare-metal computers in a local network to buid dev/work infrastructure.
Use cases:
Case 1: I use a Portainter tool to have a UI for docker environments. It allows to manage the single local docker env via unix socket and multiple ones via tcp.
Case 2: CI/CD tools (gitlab, jenkins) could (and will) be installed outside of local Macbook machine but with ability to manage containers on it.

Both cases require docker API exposed via tcp.

[abiosoft]

Now I understand.

You can modify the docker daemon settings via colima start --edit and listen on both socket and port. Kindly check the FAQ

Ensure to listen on the default socket as well, otherwise the local docker context would be broken because Colima makes the assumption.

I am limited to mobile for some days and unable to respond comprehensively.

Thanks.

[canemacchina]

what do you mean exactly by "Ensure to listen on the default socket as well"?
I also have the need to connect to the docker daemon through TCP, so I tried this config:

docker: {"hosts": ["tcp://0.0.0.0:2375", "unix:///Users/lorenzo/.colima/default/docker.sock"]}

But colima fails to start with this error:

INFO[0000] starting colima
INFO[0000] runtime: docker
INFO[0000] preparing network ...                         context=vm
INFO[0000] starting ...                                  context=vm
INFO[0021] provisioning ...                              context=docker
INFO[0021] starting ...                                  context=docker
FATA[0083] error starting docker: error running [lima sudo docker info], output: "errors pretty printing info", err: "exit status 1"

Instead, everything seems perfectly working using this daemon config

docker:  {"hosts": ["tcp://0.0.0.0:2375", "unix:///var/run/docker.sock"]}

That doesn't make sense to me because unix:///var/run/docker.sock does not exist

[abiosoft]

That doesn't make sense to me because unix:///var/run/docker.sock does not exist

@canemacchina

It exists :) but in the VM, not on your machine. The one on your machine unix:///Users/lorenzo/.colima/default/docker.sock is actually being forwarded to unix:///var/run/docker.sock in the VM.

The docker config is for the VM, so it needs to point to the socket in the VM.

[canemacchina]

ah ok perfect! Thanks!

[tobias-schuele]

As described above, I added docker: {"hosts": ["tcp://0.0.0.0:2375", "unix:///var/run/docker.sock"]} to the config. However, when starting Colima with colima start --network-address, I get the following error messages (both with 0.6.6 and HEAD):

FATA[0272] error starting docker: error running [lima sudo service docker start], output: "Warning: The unit file, source configuration file or drop-ins of docker.service changed on disk. Run 'systemctl daemon-reload' to reload units.", err: "exit status 1"

Any ideas what's going wrong? Thanks!

[stowns]

same

[matdan9]

I had the same problem. Looks like it's related to systemd not reloading the dockerd config. Deleting the colima VM colima delete and recreating it colima start --edit seems to fix the problem.