It is possible to configure an autoscaling group to use Launch Template instead of Launch Configuration. I want to use it because when I configure an autoscaling group using Launch Configuration I can not set mixed instances policy (ondemand + spot instances).
I have created launch template (name 'test') with such configuration:
AMI Ubuntu 16,
some security groups and IAM Role with EC2, AutoScaling and CodeDeploy full access
and all other settings in default *Not specified* state.
Launch Template ID lt-0e1111
Default Version Number 1
Launch Template Name test
Owner arn:aws:iam::11111111:user/itsp
Version 1 (Default)
Name test
Created by arn:aws:iam::179961111910:user/itsp
Description Not specified
Date Created February 8, 2019 at 12:47:54 AM UTC+3
AMI ID ami-0bf1924ed30c48a6a
Availability Zone Not specified
Security Groups Not specified
Instance Type t2.nano
Key pair name itsp
Security Group Ids sg-07e746cf0e4935b84,sg-0a45f7e6ac321614d,sg-27bf6d4b
Advanced Details:
Purchasing option Not specified
Spot request type Not specified
Spot interruption behavior Not specified
IAM Instance Profile Not specified
Stop - Hibernate behavior Not specified
Monitoring Not specified
EBS-optimized instance Not specified
Tenancy Host ID Not specified
RAM Disk ID Not specified
User Data Not specified
Elastic Inference Accelerator Not specified
Maximum Spot price Not specified
Spot expiry date -
Spot block duration Not specified
Shutdown behavior Not specified
Termination Protection Not specified
Placement Group Name Not specified
Tenancy Not specified
Tenancy Affinity Not specified
Kernel ID Not specified
Elastic Graphics Not specified
Then I am creating an AutoScaling group with that launch template. Here is my autoscaling group configuration:
Launch Template test
Launch Template Version 1
Launch Template Description -
Instance Types -
Spot Diversity 0
Optional On-Demand Base 0
On-Demand Percentage 0%
Desired Capacity 1
Min 1
Max 1
Availability Zone(s) eu-central-1a
Subnet(s) subnet-53075e38
Classic Load Balancers -
Target Groups -
Health Check Type EC2
Health Check Grace Period 300
Instance Protection -
Termination Policies Default
Suspended Processes -
Placement Groups -
Default Cooldown 300
Enabled Metrics -
Creation Time Fri Feb 08 00:50:16 GMT+300 2019
Service-Linked Role arn:aws:iam::11111111111:role/aws-service-role/autoscaling.amazonaws.com/AWSServiceRoleForAutoScaling
Such configuration appears to be set when I choose
*Fleet Composition - Adhere to the launch template
The launch template determines the instance type and purchase option (On-Demand or Spot).*
option and my lauch template configuration is described above.
So there are no spots in my autoscaling group.
Then I configure a CodeDeploy application 'Application' and DeploymentGroup 'DG_TEST' inside of it with such configuration:
Deployment type Blue/green
Service role ARN arn:aws:iam::11111111111111:role/CodeDeployRole
Deployment configuration CodeDeployDefault.AllAtOnce
Rollback enabled -
Environment configuration: Amazon EC2 Auto Scaling groups -> Name 'test'
Ok, after everything is set up I set desired instances count to 1 in my Autoscaling group and new on-demand instance spawns in my 'test' group.
I am now starting a blue-green deployment with AWS CLI command:
aws deploy create-deployment --application-name Application --s3-location bucket=codedeploybucket,key=application,bundleType=zip --deployment-group-name DG_TEST
And I can see new autoscaling group with same lauch template appears in my green fleet.
So we just checked all our services are working as expected.
Now I want to tweak my autoscaling group configuration a little to mix ondemand and spot instances in it:
Fleet Composition - Combine purchase options and instances
Choose a mix of On-Demand Instances and Spot Instances and multiple instance types. Spot Instances are automatically launched at the lowest price available.
Instance Types:
t2.nano (1vCPUs, 0.5GiB)
t2.small (1vCPUs, 2GiB)
t2.micro (1vCPUs, 1GiB)
Maximum Spot Price - Use default (recommended)
Spot Allocation Strategy - Diversify Spot Instances across your 2 lowest priced instance types per Availability Zone
Optional On-Demand Base - Designate the first 0 instances as On-Demand
On-Demand Percentage 50% On-Demand and 50% Spot
Desired Capacity 1
Min 1
Max 1
I save it and try same AWS CLI command
aws deploy create-deployment --application-name Application --s3-location bucket=codedeploybucket,key=application,bundleType=zip --deployment-group-name DG_TEST
And now I have my deployment failed with such comment:
The following validation error occurred: Valid requests must contain either LaunchTemplate, LaunchConfigurationName, InstanceId or MixedInstancesPolicy parameter. (Service: AmazonAutoScaling; Status Code: 400; Error Code: ValidationError; Request ID: 11111114-2b2b-11e9-93b7-11111111111)
Another strange thing is that if I configure my Launch Templae to use spots only:
Launch Template ID lt-00dc4c187d1111159
Default Version Number 1
Launch Template Name test2
Owner
arn:aws:iam::179961111910:user/itsp
Version 1 (Default)
Name test2
Created by arn:aws:iam::179961111910:user/itsp
Description Not specified
Date Created February 8, 2019 at 1:14:13 AM UTC+3
AMI ID ami-0bf1924ed30c48a6a
Availability Zone Not specified
Security Groups Not specified
Instance Type t2.micro
Key pair name itsp
Security Group Ids sg-0a45f7e6ac321614d,sg-07e746cf0e4935b84,sg-27bf6d4b
Advanced details
Purchasing option spot
Spot request type one-time
Spot interruption behavior Not specified
IAM Instance Profile CodeDeployRole (arn:aws:iam::1111111111:instance-profile/CodeDeployRole)
Stop - Hibernate behavior Not specified
Monitoring Not specified
EBS-optimized instance Not specified
Tenancy Host ID Not specified
RAM Disk ID Not specified
User Data Not specified
Elastic Inference Accelerator Not specified
Maximum Spot price Not specified
Spot expiry date -
Spot block duration Not specified
Shutdown behavior Not specified
Termination Protection Not specified
Placement Group Name Not specified
Tenancy Not specified
Tenancy Affinity Not specified
Kernel ID Not specified
Elastic Graphics Not specified
and use that option in my autoscaling group:
*Fleet Composition - Adhere to the launch template
The launch template determines the instance type and purchase option (On-Demand or Spot).*
Then after same deployment command I receive another error in CodeDeploy (failed deployment reason):
The IAM role arn:aws:iam::179966034910:role/CodeDeployRole does not give you permission to perform operations in the following AWS service: AmazonAutoScaling. Contact your AWS administrator if you need help. If you are an AWS administrator, you can grant permissions to your users or groups by creating IAM policies.
And that is definitely not the case because my role for CodeDeploy have full EC2, Autoscaling and CodeDeploy access.
Am I doing something wrong or that are bugs? How can I configure mixed autoscaling group (ondemand+spot instances) with codedeploy using launch templates? Thanks in advance!
