6.4. Server Autoscale

What is the Autoscale Command?

The Server Autoscale command can automatically start and/or stop cloud resources once configured. First the status of the scheduler is used to determine if jobs exceed the available cores or if the cores exceed the desired idle cores. Second a check is made for each of the compute nodes to ensure connectivity is still working and the node doesn’t need to be rebooted.

Once the number of nodes to start or stop has been determined calls are made to bring the cluster to the desired state. The number of always ready nodes can be set to avoid delays between checks, or set lower to reduce cost.

Arguments

Parameter name	Requirement
-h –help	optional
-port	optional
-server	optional
-test-only	optional
-log-file	optional
-otype	optional

-h

Help

-port

Port of the autoscale server. Default 10080

-server

Hostname or ip of autoscale server.

-test-only

Get a list of which servers will be brought up/taken down without running api calls to cloud provider.

-log-file

Path to write log file.

-otype

Terminal output type. If none is selected, no output will be written.

Configuration

To configure Lucy to run this command, Lucy needs a valid connection to d3VIEW and the HPC Servers will need to be available from the server running the command. The HPC Server information will determine scheduler information and the autoscale server will determine the amount of free nodes. The autoscale server needs to have access to the cloud scheduler to check the status and have the ability to run cloud service provider api calls.