ALB and ASG
Auto scalling group
ASG
is a collection of ec2 instance used for the pupose of automated scalling
asg helps us to scale out (add) or scales in (remove) ec2 instance to match load for your application
we need to define
min size - never goes below this side
max size - never goes to this size
desire size - current size
Scalling policy: 3 types
dynamic scalling:
a] target tracking
ec2 sends predefined metrics to CloudWatch eg cpu, memory
suppose u want to maintain average 50% CPU utlization on each node
create cloudwatch alarm automatically for us
b] step scaling
in target tracking we defined metrics but not define how much ec2 to provision
here define how much ec2 instance to provision when threshold match
we need to cloudwatch alarm for scalling policy
support multiple scalling adjustments as shown in image
c] simple scaling
single scaling adjustment
2] predictive scaling:
3] schedule policy:
schedule scalling based on event eg scale during black friday
Let's Proceed
Create security group
for ALB
allow http for all
and for ASG
allow all tcp for alb security group
allow ssh
Create Auto Scaling Group
step1 - select or create launch template
launch template:
A Launch Template is a resource in AWS EC2 that defines the configuration for EC2 instances.
It includes:
AMI ID (operating system + application image)
Instance type (e.g., t3.medium)
Key pair (SSH access)
Security groups
Network settings (VPC, subnet) - select at time of configuring auto scaling group
Storage (EBS volumes)
User data script (bootstrap script) - will run as root user
Step 2 - Create an Auto Scaling Group
Navigate to the Auto Scaling Groups section in the AWS Management Console.
Click on Create Auto Scaling group.
Provide a name for your Auto Scaling group.
Select the launch template you created in Step 1.
Under networking Choose the VPC and AZ, Subnet(s) where you want your group to launch instances.
Select load balancer
a] no LB
b] Existing LN
c] new LB
schema - internet or internal facing
vpc - ame as we select in networking
keep az default one
add listener and target group
enable health check for elb - tick yes
Configure group size and scaling policy
group size
desire capacity
min cap
max cap
Review and Create
suppose we have any application running on port 80
when we hit alb dns we see our app
go to auto scaling group
then auto scaling tab
insde this tab we have multiple option for scalling policy
we will see all 3 policies
dynamic scaling policy
when we select metric threshould it create 2 alarm in cloudwatch for scale in and scale out
predictive policy
schedule policy
Load testing
put load on ec2 instance by using tool called
sudo stress cpu
check cpu utilization in asg > monitoring > check graph
go to alarm in cloudwatch u will see in alarm in red on ok green for no alarm
go to activity
u will see new ec2 launch
for memory utilization
custom cloudwatch alarm
step - install agent, create cloudwatch metrics alarm eg memory threshold for ec2
assign alarm to ASG and choose steps scalling or single scaling
Good one, Pavan 👍 This is a common confusion in AWS.
🔹 Can we scale ASG on memory utilization with Target Tracking?
By default → No.
AWS Auto Scaling Groups (ASG) only support scaling on metrics from CloudWatch.
Out-of-the-box, EC2 instances do not publish memory utilization to CloudWatch.
Only CPUUtilization, NetworkIn, NetworkOut, etc. are available.
🔹 How to Scale on Memory Utilization (Workaround)
Install CloudWatch Agent
On each EC2, install and configure the CloudWatch Agent to publish
mem_used_percent(or similar custom metric).Example config (
/opt/aws/amazon-cloudwatch-agent/etc/amazon-cloudwatch-agent.json):Start the agent → metrics flow into CloudWatch under
CWAgent.
Create a CloudWatch Custom Metric Alarm
Example: Trigger when memory > 75% for 2 data points.
Attach Alarm to ASG Scaling Policy
Instead of Target Tracking, you’ll use a Step Scaling or Simple Scaling policy linked to the CloudWatch Alarm.
🔹 Why not Target Tracking directly?
Target Tracking only works with predefined metrics (like CPUUtilization) or some ALB metrics.
For custom metrics (like memory), AWS does not allow direct target tracking.
Instead, you simulate it with CloudWatch Alarms + Step Scaling.
✅ Summary
CPUUtilization → Target Tracking ✅
MemoryUtilization → Needs CloudWatch Agent → Custom Metric → CloudWatch Alarm → Step Scaling ❌ (not target tracking)
👉 Do you want me to show you a Terraform example of scaling an ASG based on high memory utilization (using custom metric + alarm)?
Last updated