That's strange as well - we'll need to try and reproduce it

but it also has the --create-queue flag but it didn’t create the queue I needed to create it manually

Ooo oops

It's not that - there's probably something wrong in the pod configuration where the services agent is running since it fails launching a docker due to a socket issue

The 3 - is actually :- followed by --

it has 3 “-” in the DAEMON_OPTIONS
PLus I don’t understand where is the repo for that image because the repo allegroai/clearml-agent has a different entrypoint file

Maybe this?
It’s the entrypoint file for the docker

You think it’s the “--services-mode” flag?

Can you point me to the direction maybe?
I already made changes in the clearml-server repo and made my own image for the helm-charts… maybe I could fix this as well

Sure no problem thanks!

Yeah, it seems this is an issue with the services agent in the helm chart - can you please open an issue in the helm repo and we'll try to handle it there?

Works when the task is running locally

I only enabled the pod and created services queue

It seems like the agent tries to spin up a docker image but fails since docker is not configured in the pod

Wait, is the clearml-agent-services configured in docker mode or not?

why locally it doesn’t try to run docker?

all the api use gunicorn

helm charts

What do you mean?

How are you running your server?

When running with --local True
everything works btw… but locally 🙃

This happens

and when I try to do this

I believe because of this

SuccessfulKoala55
This is the error

The agent has a preconfigured default docker image name

But why does it try to run “docker run”?

One moment

What was the error? You can probably see it in the task log