Skip to content
MLflow Experiment Tracker AI Platform & Data · mlflow 3.17.2 Launch on AWS

Getting started

Launch the MLflow tracking server, record one synthetic run, and open the UI through a tunnel.

You need: an AWS account that can subscribe on AWS Marketplace, launch EC2, and start Session Manager. To open the UI from your workstation, install the AWS CLI and the Session Manager plugin. An SSH tunnel is the other path the published guide names.

Watch on YouTube.

1. Subscribe

Open the MLflow Experiment Tracker listing, choose View purchase options, review the terms, and subscribe. When the subscription is active, choose Launch your software.

2. Launch the instance

On the EC2 launch page:

  1. Choose t3.medium (4 GiB of memory) or larger.
  2. Under Configure storage, use an encrypted gp3 root volume of at least 32 GiB.
  3. Require IMDSv2.
  4. Attach an IAM instance profile that includes AmazonSSMManagedInstanceCore if you use Session Manager.
  5. Leave inbound application ports closed.
  6. Choose Launch instance.

Allow about two minutes for first boot to create the accounts and start the service.

3. Connect

Session Manager starts as ssm-user. In the EC2 console, choose Connect, then Session Manager, then Connect. Or:

aws ssm start-session --target <instance-id>

Wait for cloud-init, then run the walkthrough as ubuntu:

sudo cloud-init status --wait
sudo -iu ubuntu bash /opt/sushi/docs/demo.sh

4. Run the walkthrough

The demo creates one synthetic experiment, logs alpha=0.25 and score=0.125, uploads a text artifact, and reads the artifact bytes and their SHA-256 back. Results are grouped under ~/mlflow-results/. Keep the manifest path it prints.

The helper reads ~/.config/sushi/mlflow/access.json for you. That file is created on first boot and is readable only by ubuntu. Open it in your session editor when you need the browser sign-in. Do not copy it into shared logs.

A second identity, /etc/mlflow/owner.json, is the administrator account. Read it only through root. The service receives the administrator password during account bootstrap. Accounts, the CSRF key, and both databases are generated after launch and kept across reboots.

5. Open the tracking UI

Forward workstation port 5000 to 127.0.0.1:5000 on the instance with Session Manager or an SSH tunnel. Then open http://localhost:5000. Leave the HTTP endpoint behind the tunnel.

aws ssm start-session --target <instance-id> \
  --document-name AWS-StartPortForwardingSession \
  --parameters '{"portNumber":["5000"],"localPortNumber":["5000"]}'

Sign in with the customer account from access.json.

6. Read the saved result

sushi-mlflow read creates no new run. Use it after a restart or a reboot. Replace the directory name with the one the demo printed.

sushi-mlflow read ~/mlflow-results/RESULT-DIRECTORY/manifest.json

Uploaded artifacts are stored files. This server does not serve uploaded models or execute pickles. Next, see Configuration.

Last updated October 27, 2026 · Suggest an improvement