Launch AlwaysOn nodes

Why AlwaysOn instances?

By default, Engineering Development Hub provisions on-demand capacity when there are jobs in the normal queue. This means any job submitted will wait in the queue 5 to 8 minutes until EC2 capacity is ready.

If you want to avoid this penalty, you can provision "AlwaysOn instances" (or pre-bake an on-demand AMI). Please note you will be charged until you manually terminate it or specify --terminate_when_idle option.

How to launch an AlwaysOn instances?

./edhctl nodes create-always-on  --scheduler-identifier openpbs-default-soca-te61 \
  --job-owner mcrozes \
  --job-queue alwayson \
  --instance-type c6i.xlarge \
  --nodes 2 --base-os amazonlinux2023 \
  --instance-ami ami-025ca978d4c1d9825

Capacity tagged alwayson-5d4a6513-fffc-4c24-bc69-ee7b1d762390 has been created. This capacity will run 24/7 until you delete the associated Cloudformation stack.

./edhctl config get --key "/configuration/Schedulers/" --output json
{
    "/configuration/Schedulers/openpbs-default-soca-te61": {
        "enabled": true,
        "provider": "openpbs",
        "endpoint": "ip-11-0-64-70.us-east-2.compute.internal",
        "binary_folder_paths": "/opt/edh/soca-te61/schedulers/default/pbs/bin",
        "soca_managed_nodes_provisioning": true,
        "identifier": "openpbs-default-soca-te61",
        "pbs_configuration": {
            "install_prefix_path": "/opt/edh/soca-te61/schedulers/default/pbs",
            "pbs_home": "/opt/edh/soca-te61/schedulers/default/pbs/var/spool/pbs"
        }
    }
}

./edhctl nodes create-always-on --help
Usage: app.py nodes create-always-on [OPTIONS]

Options:
  --scheduler-identifier TEXT     Create a new always on for a specific
                                  [required]
  --job-owner TEXT                Job owner for the new always on node(s)
                                  [required]
  --job-queue TEXT                Job Queue  [required]
  --instance-type TEXT            Instance Type to provision  [required]
  --nodes INTEGER                 Number of nodes to create  [required]
  --instance-ami TEXT             EC2 Image ID for the node(s)  [required]
  --base-os [amazonlinux2|amazonlinux2023|rhel8|rhel9|rocky8|rocky9|ubuntu2204|ubuntu2404]
                                  Operating system to use  [required]
  --root-size INTEGER             Size of the root partition. Will use the
                                  default one configured for your AMI if not
                                  set
  --instance-profile TEXT         IAM Instance Profile for the node(s)
  --security-groups TEXT          Security group(s) to assign
  --subnet-id TEXT                Subnet(s) to deploy capacity on
  --capacity-reservation-id TEXT  Capacity Reservation ID to use
  --anonymous-metrics / --no-anonymous-metrics
                                  Enable or disable anonymous data tracking
  --fsx-lustre TEXT               FSx for Lustre association: True for new
                                  FSxL or provide fs-id
  --fsx-lustre-size INTEGER       FSx for Lustre size if used
  --fsx-lustre-deployment-type TEXT
                                  FSx for Lustre deployment type
  --fsx-lustre-per-unit-throughput INTEGER
                                  FSxL throughput per unit
  --fsx-lustre-storage-type TEXT  FSxL storage type
  --scratch-iops INTEGER          Use io2 for scratch instead of gp3 (IOPS)
  --scratch-size INTEGER          Custom /scratch size in GiB
  --spot-price TEXT               Spot price (float, int, or 'auto')
  --spot-allocation-count INTEGER
                                  Spot allocation count
  --spot-allocation-strategy [capacity-optimized|lowest-price|diversified]
                                  Spot allocation strategy
  --keep-ebs / --no-keep-ebs      Preserve EBS after capacity deletion
  --placement-group / --no-placement-group
                                  Enable placement group for the node(s)
  --efa-support / --no-efa-support
                                  Enable Elastic Fabric Adapter (EFA)
  --force-ri / --no-force-ri      Require Reserved Instance only
  --ht-support / --no-ht-support  Enable or disable Hyper-Threading
  --help                          Show this message and exit.
  ```



List all Always On instance
```bash

# This command will return all EC2 instance information
./edhctl nodes list --node-lifecycles alwayson

# You can filter with jq to only get specific info such as the Private DNS Name of all your Always On instances
./edhctl nodes list --node-lifecycles alwayson | jq -r '.[].PrivateDnsName'
ip-11-0-64-70.us-east-2.compute.internal
ip-11-0-250-76.us-east-2.compute.internal
ip-11-0-243-250.us-east-2.compute.internal
ip-11-0-253-157.us-east-2.compute.internal