A demo with variants showing fleet creation with a Spot and On-Demand combination, fair instances spread over multi-AZ and, optionnaly, with vertical scaling configuration.
WARNING: Do not reuse AS-IS the generated template for real world projects as instances
are customized with a cgi-bin named 'cpu' to generate CPU load for demo purposes. Embedding such
cpu-stress-tool cgi-bin in real project by mistake could offer an easy path to DDoS for malicious people.
WARNING: When Running the whole fleet at full size, it can cost up to 50$ per day! When running at minimum size, the fleet cost is about 5$ per day.
The script file named 'deploy-test-instances.sh' generates a Cloudformation template and deploys it directly.
By default, it defines a main autoscaled fleet of 20 instances with 3 x Spot t3.medium, 4 x Spot c5.large,
8 x On-Demand c5.large and 5 x On-Demand c5.xlarge.
FLEET_SPECIFICATION=${FLEET_SPECIFICATION:-"t3.medium,Spot=True,Count=3;c5.large,Spot=True,Count=4;c5.large,Count=8;c5.xlarge,Count=5"}The script also generates 3 'subfleets'. There are used by the demo-scheduled-events.
MySubfleet1:- 2 x t3.micro Spot,
- 2 x t3.micro On-Demand,
- 1 x RDS MySQL
MySubfleet2:- 2 x t3.micro On-Demand
MySubfleet3:- 2 x Aurora DB
Note: You must launch demo deployment from the CloneSquad DevKit Docker container! See instructions..
./deploy-test-instances.shVertical scaling can be configured to define priorities between instance types of the fleet.
The below command line will inject the configure-ligthhouse-instance.yaml policy definition in the Configuration DynamoDB table. Please look at comments in this YAML file to understand what is the vertical scaling policy defined.
${CLONESQUAD_DIR}/tools/cs-kvtable CloneSquad-${GroupName}-Configuration import <configure-ligthhouse-instance.yaml- When existing persistent Spot instances are updated by the template.yaml Cloudformation scripts, former Spot request are not currectly cancelled (CloudFormation issue?). User need to cancel it by himself (console or API).