allegroai / Clearml Server

ClearML - Auto-Magical Suite of tools to streamline your ML workflow Experiment Manager, ML-Ops and Data-Management

In v0.16, the Elasticsearch subsystem of ClearML Server has been upgraded from version 5.6 to version 7.6. This change necessitates the migration of the database contents to accommodate the change in index structure across the different versions.

Follow this procedure to migrate existing data.

ClearML Server

Formerly known as Trains Server

The ClearML Server is the backend service infrastructure for ClearML. It allows multiple users to collaborate and manage their experiments. By default, ClearML is set up to work with the ClearML demo server, which is open to anyone and resets periodically. In order to host your own server, you will need to launch the ClearML Server and point ClearML to it.

The ClearML Server contains the following components:

• The ClearML Web-App, a single-page UI for experiment management and browsing
• RESTful API for:
  • Documenting and logging experiment information, statistics and results
  • Querying experiments history, logs and results
• Locally-hosted file server for storing images and models making them easily accessible using the Web-App

You can quickly deploy your ClearML Server using Docker, AWS EC2 AMI, or Kubernetes.

System design

The ClearML Server has two supported configurations:

• Single IP (domain) with the following open ports

  • Web application on port 8080
  • API service on port 8008
  • File storage service on port 8081

• Sub-Domain configuration with default http/s ports (80 or 443)

  • Web application on sub-domain: app.*.*
  • API service on sub-domain: api.*.*
  • File storage service on sub-domain: files.*.*

Launching The ClearML Server

Prerequisites

The ports 8080/8081/8008 must be available for the ClearML Server services.

For example, to see if port 8080 is in use:

• Linux or macOS:

    sudo lsof -Pn -i4 | grep :8080 | grep LISTEN
  

• Windows:

    netstat -an |find /i "8080"
  

Launching

Launch The ClearML Server in any of the following formats:

• Pre-built AWS EC2 AMI
• Pre-built GCP Custom Image
• Pre-built Docker Image
  • Linux
  • macOS
  • Windows 10
• Kubernetes
  • Kubernetes Helm
  • Manual Kubernetes installation

Connecting ClearML to your ClearML Server

By default, the ClearML client is set up to work with the ClearML demo server.
To have the ClearML client use your ClearML Server instead:

• Run the clearml-init command for an interactive setup.

• Or manually edit ~/clearml.conf file, making sure the server settings (api_server, web_server, file_server) are configured correctly, for example:

    api {
        # API server on port 8008
        api_server: "http://localhost:8008"
  
        # web_server on port 8080
        web_server: "http://localhost:8080"
  
        # file server on port 8081
        files_server: "http://localhost:8081"
    }
  

Note: If you have set up your ClearML Server in a sub-domain configuration, then there is no need to specify a port number, it will be inferred from the http/s scheme.

After launching the ClearML Server and configuring the ClearML client to use the ClearML Server, you can use ClearML in your experiments and view them in your ClearML Server web server, for example http://localhost:8080.
For more information about the ClearML client, see ClearML.

ClearML-Agent Services

As of version 0.15 of ClearML Server, dockerized deployment includes a ClearML-Agent Services container running as part of the docker container collection.

ClearML-Agent Services is an extension of ClearML-Agent that provides the ability to launch long-lasting jobs that previously had to be executed on local / dedicated machines. It allows a single agent to launch multiple dockers (Tasks) for different use cases. To name a few use cases, auto-scaler service (spinning instances when the need arises and the budget allows), Controllers (Implementing pipelines and more sophisticated DevOps logic), Optimizer (such as Hyper-parameter Optimization or sweeping), and Application (such as interactive Bokeh apps for increased data transparency)

ClearML-Agent Services container will spin any task enqueued into the dedicated services queue. Every task launched by ClearML-Agent Services will be registered as a new node in the system, providing tracking and transparency capabilities.
You can also run the ClearML-Agent Services manually, see details in ClearML-agent services mode

Note: It is the user's responsibility to make sure the proper tasks are pushed into the services queue. Do not enqueue training / inference tasks into the services queue, as it will put unnecessary load on the server.

Advanced Functionality

The ClearML Server provides a few additional useful features, which can be manually enabled:

• Web login authentication
• Non-responsive experiments watchdog

Restarting ClearML Server

To restart the ClearML Server, you must first stop the containers, and then restart them.

docker-compose down
docker-compose -f docker-compose.yml up

Upgrading

ClearML Server releases are also reflected in the docker compose configuration file.
We strongly encourage you to keep your ClearML Server up to date, by keeping up with the current release.

Note: The following upgrade instructions use the Linux OS as an example.

To upgrade your existing ClearML Server deployment:

1. Shut down the docker containers

   docker-compose down
   

2. We highly recommend backing up your data directory before upgrading.

   Assuming your data directory is /opt/clearml, to archive all data into ~/clearml_backup.tgz execute:

   sudo tar czvf ~/clearml_backup.tgz /opt/clearml/data
   

   Restore instructions:

   To restore this example backup, execute:

   sudo rm -R /opt/clearml/data
   sudo tar -xzf ~/clearml_backup.tgz -C /opt/clearml/data
   

3. Download the latest docker-compose.yml file.

   curl https://raw.githubusercontent.com/allegroai/trains-server/master/docker/docker-compose.yml -o docker-compose.yml 
   

4. Configure the ClearML-Agent Services (not supported on Windows installation). If TRAINS_HOST_IP is not provided, ClearML-Agent Services will use the external public address of the ClearML Server. If TRAINS_AGENT_GIT_USER / TRAINS_AGENT_GIT_PASS are not provided, the ClearML-Agent Services will not be able to access any private repositories for running service tasks.

   export TRAINS_HOST_IP=server_host_ip_here
   export TRAINS_AGENT_GIT_USER=git_username_here
   export TRAINS_AGENT_GIT_PASS=git_password_here
   

5. Spin up the docker containers, it will automatically pull the latest ClearML Server build

   docker-compose -f docker-compose.yml pull
   docker-compose -f docker-compose.yml up
   

* If something went wrong along the way, check our FAQ: Common Docker Upgrade Errors.

Community & Support

If you have any questions, look to the ClearML FAQ, or tag your questions on stackoverflow with 'clearml' tag.

For feature requests or bug reports, please use GitHub issues.

Additionally, you can always find us at [email protected]

License

Server Side Public License v1.0

The ClearML Server relies on both MongoDB and ElasticSearch. With the recent changes in both MongoDB's and ElasticSearch's OSS license, we feel it is our responsibility as a member of the community to support the projects we love and cherish. We believe the cause for the license change in both cases is more than just, and chose SSPL because it is the more general and flexible of the two licenses.

This is our way to say - we support you guys!