Skip to content

Getting Started

To use the MMBatch Engine with AWS Batch you need to deploy the MMBatch Managmenet Server (MMS) and ensure that your worker nodes fetch and install the agent and runtime at startup.

You can do this manually or using a CloudFormation template which we provide (we need to allowlist your AccountID).

MMBatch Management Server

How to install the MMS comes down to installing the package nativly or as docker container. After installation a service will be started on the host.

You can change the configuration using the API itself.

Change Port

You can change the port from https to http with this command:

curl -skX PUT https://127.0.0.1/api/v1/configKV \
-H "Content-Type: application/json" \
-d "{\"kvMap\": {\"addr\":\"http://0.0.0.0:8080\"}}" |jq .

# curl -skX PUT https://0.0.0.0/api/v1/configKV \
    -H "Content-Type: application/json"   \
    -d "{\"kvMap\": {\"addr\":\"http://0.0.0.0:8080\"}}" |jq .
{
"addr": "http://0.0.0.0:8080",
"staticFolder": "mmabWeb",
"log": {
    "file": "/var/log/memverge/mmab.log",
    "level": "info",
    "maxSizeMB": 10,
    "maxBackups": 10
},
"security": {
    "certFile": "/etc/memverge/mmab/server.crt",
    "keyFile": "/etc/memverge/mmab/server.pem",
    "cognito": {
    "enabled": false,
    "userPoolID": "",
    "identityPoolID": "",
    "clientID": ""
    }
},
"ckpt": {
    "ckptMode": "unknown",
    "ckptImagePath": "",
    "ckptInterval": 900000000000,
    "ckptFiles": null,
    "IRMapScanPaths": null,
    "ckptOnSigTerm": false,
    "rootFSDiff": true
},
"node": {
    "heartbeat": 30000000000,
    "ttl": 300000000000
}
}

Enable Checkpoints

After you changed the port (or not) you can enable checkpoints for all connected agents.

curl -sk -X PUT http://localhost:8080/api/v1/ckptConfig \
    -H "Content-Type: application/json" \
    -d '{"ckptMode":"iterative","ckptImagePath":"/mmc-checkpoint","ckptInterval":120000000000,"rootFSDiff":true,"diagnosisMode":true,"ckptOnSigTerm":true}'|jq .

curl -sk -X PUT https://localhost/api/v1/ckptConfig \
    -H "Content-Type: application/json" \
    -d '{"ckptMode":"iterative","ckptImagePath":"/mmc-checkpoint","ckptInterval":120000000000,"rootFSDiff":true,"diagnosisMode":true,"ckptOnSigTerm":true}'|jq .

$ curl -sk -X PUT http://localhost:8080/api/v1/ckptConfig \
    -H "Content-Type: application/json" \
    -d '{"ckptMode":"iterative","ckptImagePath":"/mmc-checkpoint","ckptInterval":120000000000,"rootFSDiff":true,"diagnosisMode":true,"ckptOnSigTerm":true}'|jq .
{
"ckptMode": "iterative",
"ckptImagePath": "/mmc-checkpoint",
"ckptInterval": 120000000000,
"ckptFiles": null,
"IRMapScanPaths": null,
"ckptOnSigTerm": true,
"diagnosisMode": true,
"rootFSDiff": true
}

Shared Checkpoint Filesystem

For this little quickstart we use an EFS file-system. You'd need to create a filestem in the same VPC you host your Compute Environment in.

Suggested Filesystem is JuiceFS

EFS is easy to install and it makes for a short UserData in this quick start. Howerver, we recommend JuiceFS as the default checkpoint filesystem.

Install Agent

To install the agent, you need to change the include the UserData into the launch-template which is used within the Compute Enviroment. While we are at it we'll also install awscli which is needed for Nextflow runs.

MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="//"

--//
Content-Type: text/cloud-config; charset="us-ascii"

#cloud-config
write_files:
- path: /root/mm-install.sh
  permissions: 0744
  owner: root
  content: |
    #!/bin/bash
    # Install required dependencies
    yum install -y amazon-efs-utils bzip2 wget unzip fuse gcc python3 bash jq
    # Define Miniconda URL and installation path
    MINICONDA_URL="https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh"
    INSTALL_PATH="/nextflow_awscli"
    # Function to install Miniconda and AWS CLI
    install_aws_cli() {
        echo "Starting AWS CLI installation..."
        wget -q $MINICONDA_URL -O Miniconda3-latest-Linux-x86_64.sh
        bash Miniconda3-latest-Linux-x86_64.sh -b -f -p $INSTALL_PATH
        $INSTALL_PATH/bin/conda install -c conda-forge -y awscli
        rm -f Miniconda3-latest-Linux-x86_64.sh
    }
    # Attempt AWS CLI installation
    install_aws_cli

    # Retry if installation fails
    if ! $INSTALL_PATH/bin/aws --version &> /dev/null; then
        echo "AWS CLI installation failed. Retrying..."
        install_aws_cli
        if ! $INSTALL_PATH/bin/aws --version &> /dev/null; then
            echo "AWS CLI installation failed after retry. Exiting."
            exit 1
        fi
    fi

    echo "AWS CLI installed successfully."

    mkdir -p /mmc-checkpoint
    mount -t efs -o tls fs-03ae6723a107f8310:/ /mmc-checkpoint
    # Install MM batch engine
    curl -k http://18.143.94.11:8080/api/v1/scripts/install-pagent | bash
    systemctl restart docker || true
    systemctl restart --no-block ecs
    echo "Setup complete."

runcmd:
  - bash /root/mm-install.sh

Create CE

Please keep in mind that a Compute Environment (CE) creates a copy of the Launch Template it defined at creation time. If you alter an existing Launch Template you need to clone the CE with the new Launch Template and alter the Job Queue to pick up the new CE.