I'm here back to say that I was able to bring up clearML w/o docker. I told the lab to test it in order to see if it functions properly.
The process was:
creating the web interface by cloning clearml-web and building it. during the build I had to fix some paths, so I did and create a pull request to the original repository. I setup the apiBaseUrl and fileBaseUrl according to the location I use in the nginx.conf running the api server and the files server using uwsgi. In order to use other path from /opt/clearml I mad a little hack in the apiserver/config/basic.py and added another path for configs. running nginx server with uwsgi_pass to the socket files I used for the api and files servers (I prefer file socket over high ports)
Running redis, mongodb and elasticsearch locally made it easier since I didn't have to change the configuration of those. I guess I could use some environment variable on the uwsgi files.
Thanks you all
Hi CheekyToad28 , any specific reason you can't use the prebuilt docker images?
Also, you can simply install a VM and install the docker image inside...
Hi TimelyPenguin76
The only option for Linux is prebuilt docker image, which I cannot use on our system.
Hi CheekyToad28 , nice job 🙂
In order to use other path from /opt/clearml I mad a little hack in the apiserver/config/basic.py and added another path for configs.
This can be achieved using the CLEARML_CONFIG_DIR env var
I guess I could use some environment variable on the uwsgi files.
Out of curiosity, what would you need that var for, exactly, and what would it control?
Well, installing t manually would require a lot of work, including building the angular app, and installing various requirements for the different servers (apiserver, fileserver etc.) - without docker, each should be run in it's own virtual python environment
Hi SuccessfulKoala55
In case the code is using environment variable in runtime (such as CLEARML_CONFIG_DIR), I have to pass them on runtime. I can use "export" if I run it manually, or setup environment on the systemd file or setting them on the uwsgi configuration file.
Since I want it to run automatically when the system starts I have to set them on one of the service's configuration file.
You can try using the GCP VMDK, perhaps
Hi CheekyToad28 ,
None of the options https://clear.ml/docs/latest/docs/deploying_clearml/clearml_server#deployment works for you?
Couldn't you simply prefix the command running at system startup with the env var, i.e. CLEARML_CONFIG_DIR=/my/path the-command-to-run ?
SuccessfulKoala55 we are not using docker for security reasons. I'm trying to find a way to run the service using our system (it should run on our network) and not a stand-alone server on a VM which will need extra maintenance.