1
Q
EC2 Changing Instance Type
A
- This only works for EBS backed instances
- Stop the instance
- Instance Settings => Change Instance Type
- Start Instance
2
Q
EC2 Enhanced Networking (SR-IOV)
A
- Higher bandwidth, higher PPS (packet per second), lower latency
- Option 1: Elastic Network Adapter (ENA) up to 100 Gbps
- Option 2: Intel 82599 VF up to 10 Gbps – LEGACY
- Works for newer generation EC2 Instances
3
Q
Enhanced Networking - Elastic Fabric Adapter (EFA)
A
- Improved ENA for HPC, only works for Linux
- Great for inter-node communications, tightly coupled workloads
- Leverages Message Passing Interface (MPI) standard
- Bypasses the underlying Linux OS to provide low-latency, reliable transport
4
Q
Placement Groups
A
- Sometimes you want control over the EC2 Instance placement strategy
- That strategy can be defined using placement groups
- When you create a placement group, you specify one of the following
strategies for the group: - Cluster—clusters instances into a low-latency group in a single Availability Zone
- Spread—spreads instances across underlying hardware (max 7 instances per group per AZ) – critical applications
- Partition—spreads instances across many different partitions (which rely on different sets of racks) within an AZ. Scales to 100s of EC2 instances per group
(Hadoop, Cassandra, Kafka)
5
Q
Placement Groups Cluster
A
Pros: Great network (10 Gbps bandwidth between instances with Enhanced Networking enabled - recommended)
* Cons: If the AZ fails, all instances fails at the same time
* Use case:
* Big Data job that needs to complete fast
* Application that needs extremely low latency and high network throughput
6
Q
Placement Groups Spread
A
- Pros:
- Can span across Availability Zones (AZ)
- Reduced risk is simultaneous
failure - EC2 Instances are on different
physical hardware
Cons:
* Limited to 7 instance
Use case:
EC2
* Application that needs to maximize high availability
Hardware 6
* Critical Applications where each instance must be isolated
from failure from each other
7
Q
Placements Groups Partition
A
- Up to 7 partitions per AZ
- Can span across multiple AZs in the same region
- Up to 100s of EC2 instances
- The instances in a partition do not share racks with the instances in the EC2 other partitions
- A partition failure can affect many EC2 but won’t affect other partitions EC2
- EC2 instances get access to the partition information as metadata.
- Use cases: HDFS, HBase, Cassandra, Kafka
8
Q
Shutdown Behavior
A
- Shutdown Behavior: How should the instance
react when shutdown is done using the OS? - Stop (default)
- Terminate
- This is not applicable when shutting down from AWS console.
- CLI Attribute:
InstanceInitiatedShutdownBehavior
9
Q
Termination Protection
A
- Enable termination protection:
To protect against accidental termination in AWS Console or CLI - Exam Tip:
- We have an instance where shutdown behavior = terminate and enable terminate protection is ticked
- We shutdown the instance from the OS, what will happen ?
- The instance will still be terminated!
10
Q
InstanceLimitExceeded
EC2 Launch Tshooting - InstanceLimitExceeded
A
- # InstanceLimitExceeded: if you get this error, it means that you have reached your limit of max number of vCPUs per region
- On-Demand instance limits are set on a per-region basis
- Resolution: Either launch the instance in a different region or request AWS to increase your limit of the region
- NOTE: vCPU-based limits only apply to running On-Demand instances and Spot instances
11
Q
EC2 Launch Troubleshooting - InsufficientInstanceCapacity
A
- # InsufficientInstanceCapacity : if you get this error, it means AWS does not have that enough On-Demand capacity in the particular AZ where the instance is launched.
- Resolution :
- Wait for few mins before requesting again.
- If requesting more than 1 requests, break down the requests. If you need 5 instances, rather than a single request of 5, request one by one.
- If urgent, submit a request for a different instance type now, which can be
resized later. - Also, can request the EC2 instance in a different AZ
12
Q
EC2 Launch Troubleshooting - Instance Terminates Immediately
A
- # Instance Terminates Immediately (goes from pending to terminated)
- You’ve reached your EBS volume limit.
- An EBS snapshot is corrupt.
- The root EBS volume is encrypted and you do not have permissions to access the KMS key for decryption.
- The instance store-backed AMI that you used to launch the instance is missing a required part (an image.part.xx file).
- To find the exact reason, check out the EC2 console of AWS - instances - Description tab, note the reason next to the State transition reason label.
13
Q
EC2 SSH troubleshooting - Logging in Errors
A
- Make sure the private key (pem file) on your linux machine has 400 permissions, else you will get “Unprotected private key file” error
- Make sure the username for the OS is given Correctly when logging in via SSH. Else you will get a “Host Key Not Found”, “Permission Denied” , or “Connection Closed by [instance}” port 22 error.
14
Q
EC2 SSH troubleshooting - Connection timed out
A
- Possible reasons for “Connection timed out” to EC2 instance via SSH:
- SG is not configured correctly
- NACL is not configured correctly
- Check the route table for the subnet (routes traffic destined outside VPC to IGW)
- Instance doesn’t have a public IPv4
- CPU load of the instance is high
15
Q
SSH vs. EC2 Instance Connect
A
16
Q
EC2 Instances Purchasing Options
A
- On-Demand Instances – short workload, predictable pricing, pay by second
- Reserved (1 & 3 years)
- Reserved Instances – long workloads
- Convertible Reserved Instances – long workloads with flexible instances
- Savings Plans (1 & 3 years) –commitment to an amount of usage, long workload
- Spot Instances – short workloads, cheap, can lose instances (less reliable)
- Dedicated Hosts – book an entire physical server, control instance placement
- Dedicated Instances – no other customers will share your hardware
- Capacity Reservations – reserve capacity in a specific AZ for any duration
17
Q
EC2 On Demand
A
- Pay for what you use:
- Linux or Windows - billing per second, after the first minute
- All other operating systems - billing per hour
- Has the highest cost but no upfront payment
- No long-term commitment
- Recommended for short-term and un-interrupted workloads, where you can’t predict how the application will behave
18
Q
EC2 Savings Plans
A
- Get a discount based on long-term usage (up to 72% - same as RIs)
- Commit to a certain type of usage ($10/hour for 1 or 3 years)
- Usage beyond EC2 Savings Plans is billed at the On-Demand price
- Locked to a specific instance family & AWS region (e.g., M5 in us-east-1)
- Flexible across:
- Instance Size (e.g., m5.xlarge, m5.2xlarge)
- OS (e.g., Linux, Windows)
- Tenancy (Host, Dedicated, Default)
19
Q
EC2 Spot Instances
A
- Can get a discount of up to 90% compared to On-demand
- Instances that you can “lose” at any point of time if your max price is less than the current spot price
- The MOST cost-efficient instances in AWS
- Useful for workloads that are resilient to failure -
- Batch jobs
- Data analysis
- Image processing
- Any distributed workloads
- Workloads with a flexible start and end time
- Not suitable for critical jobs or databases
20
Q
EC2 Dedicated Hosts
A
- A physical server with EC2 instance capacity fully dedicated to your use
- Allows you address compliance requirements and use your existing server-
bound software licenses (per-socket, per-core, pe—VM software licenses) - Purchasing Options:
- On-demand – pay per second for active Dedicated Host
- Reserved - 1 or 3 years (No Upfront, Partial Upfront, All Upfront)
- The most expensive option
- Useful for software that have complicated licensing model (BYOL – Bring Your
Own License) - Or for companies that have strong regulatory or compliance needs
21
Q
EC2 Dedicated Instances
A
- Instances run on hardware that’s dedicated to you
- May share hardware with other instances in same account
- No control over instance placement (can move hardware after Stop / Start)
22
Q
EC2 Capacity Reservations
A
- Reserve On-Demand instances capacity in a specific AZ for any
duration - You always have access to EC2 capacity when you need it
- No time commitment (create/cancel anytime), no billing discounts
- Combine with Regional Reserved Instances and Savings Plans to benefit
from billing discounts - You’re charged at On-Demand rate whether you run instances or not
- Suitable for short-term, uninterrupted workloads that needs to be in a specific AZ
23
Q
Price Comparison
Example – m4.large – us-east-1
A
24
Q
AWS charges for IPv4 addresses
A
- Starting February 1st 2024, there’s a charge for all Public IPv4 created in your account
- $0.005 per hour of Public IPv4 (~ $3.6 per month)
- For new accounts in AWS, you have a free tier for the EC2 service: 750 hours of Public IPv4 per month for the first 12 months
- For all other services there is no free tier
25
AWS charges for IPv6 addresses
* What about IPv6?
* Unfortunately, many Internet Service Provider (ISP) around the world don’t
support IPv6, so the course would not work for some of you
* You can test IPv6 by going to https://test-ipv6.com/
* If you use IPv6 in this course, you’re on your own (security groups,
networking…) but you can do it!
* How to troubleshoot charges?
* Go into your AWS Bill
* Look into the AWS Public IP Insights service
* Nice article here:
https://repost.aws/articles/ARknH_OR0cTvqoTfJrVGaB8A/why-am-i-seeing-charges-for-public-ipv4-addresses-when-i-am-under-the-aws-free-tier
26
EC2 Spot Instance Requests
* Can get a discount of up to 90% compared to On-demand
* Define max spot price and get the instance while current spot price < max
* The hourly spot price varies based on offer and capacity
* If the current spot price > your max price you can choose to stop or terminate your instance
with a 2 minutes grace period.
* Other strategy: Spot Block
* “block” spot instance during a specified time frame (1 to 6 hours) without interruptions
* In rare situations, the instance may be reclaimed
* Used for batch jobs, data analysis, or workloads that are resilient to failures.
* Not great for critical jobs or databases
27
How to terminate Spot Instances?
see attachment
Note : You can only cancel Spot Instance requests that are open, active, or disabled.
Cancelling a Spot Request does not terminate instances You must first cancel a Spot Request, and then terminate the associated Spot Instances
28
Spot Fleets
* Spot Fleets = set of Spot Instances + (optional) On-Demand Instances
* The Spot Fleet will try to meet the target capacity with price constraints-
* Define possible launch pools: instance type (m5.large), OS, Availability Zone
* Can have multiple launch pools, so that the fleet can choose
* Spot Fleet stops launching instances when reaching capacity or max cost
* Strategies to allocate Spot Instances:
* lowestPrice: from the pool with the lowest price (cost optimization, short workload)
* diversified: distributed across all pools (great for availability, long workloads)
* capacityOptimized: pool with the optimal capacity for the number of instances
* priceCapacityOptimized (recommended): pools with highest capacity available, then select the pool with the lowest price (best choice for most workloads)
* Spot Fleets allow us to automatically request Spot Instances with the lowest price
29
Burstable Instances (T2/T3)
* AWS has the concept of burstable instances (T2/T3 machines)
* Burst means that overall, the instance has OK CPU performance.
* When the machine needs to process something unexpected (a spike in
load for example), it can burst, and CPU can be VERY good.
* If the machine bursts, it utilizes “burst credits”
* If all the credits are gone, the CPU becomes BAD
* If the machine stops bursting, credits are accumulated over time
30
Burstable Instances (T2/T3) cont'
* Burstable instances can be amazing to handle unexpected traffic and getting the insurance that it will be handled correctly
* If your instance consistently runs low on credit, you need to move to a different kind of non-burstable instance
31
CPU Credits
32
What happens when credit are exhausted?
* Experiment: run a CPU stress command (to peak at 100%)
* After the credits are exhausted, the measured CPU utilization drops
33
T2/T3 Unlimited
* It is possible to have an “unlimited burst
credit balance”
* You pay extra money if you go over your
credit balance, but you don’t lose in
performance
* If average CPU usage over a 24-hour period
exceeds the baseline, the instance is billed
for additional usage per vCPU/hour
* Be careful, costs could go high if you’re not
monitoring the CPU health of your
instances
34
Elastic IPs
* When you stop and then start an EC2 instance, it changes its public IP
* If you need to have a fixed public IP, you need an Elastic IP
* An Elastic IP is a public IPv4 you own as long as you don’t delete it
* You can attach it to one instance at a time
* You can remap it across instances
* You don’t pay for the Elastic IP if it’s attached to a server
* You pay for the Elastic IP if it’s not attached to a server
* With an Elastic IP address, you can mask the failure of an instance or software
by rapidly remapping the address to another instance in your account.
* You can only have 5 Elastic IP in your account (you can ask AWS to increase
that).
35
How you can avoid using Elastic IP?
* Always think if other alternatives are available to you
* You could use a random public IP and register a DNS name to it
* Or use a Load Balancer with a static hostname
36
CloudWatch Metrics for EC2
AWS Provided metrics (AWS pushes them):
* Basic Monitoring (default): metrics are collected at a 5 minute internal
* Detailed Monitoring (paid): metrics are collected at a 1 minute interval
* Includes CPU, Network, Disk and Status Check Metrics
Custom metric (yours to push):
* Basic Resolution: 1 minute resolution
* High Resolution: all the way to 1 second resolution
* Include RAM, application level metrics
* Make sure the IAM permissions on the EC2 instance role are correct !
37
EC2 included metrics
* CPU: CPU Utilization + Credit Usage / Balance
* Network: Network In / Out
* Status Check:
* Instance status = check the EC2 VM
* System status = check the underlying hardware
* Attached EBS status = check attached EBS volumes
Note - * Disk: Read / Write for Ops / Bytes (only for instance store)
* RAM is NOT included in the AWS EC2 metrics
38
Unified CloudWatch Agent
* For virtual servers (EC2 instances, on-premises servers,…)
* Collect additional system-level metrics such as RAM, processes, used disk space, etc.
* Collect logs to send to CloudWatch Logs
- No logs from inside your EC2 instance will be sent to cloud watch logs without using an Agent.
* Centralized configuration using SSM Parameter Store
* Make sure IAM permissions are correct
* Default namespace for metrics collected by Unified CloudWatch Agent is CWAgent (Can be configured and changed)
39
Unified CloudWatch Agent – procstat Plugin
* Collect metrics and monitor system utilization of individual processes
* Supports both Linux and Windows servers
* Example: amount of time the process uses CPU, amount of memory the
process uses, …
* Select which processes to monitor by:
* pid_file: name of process identification number (PID) files they create
* exe: process name that match string you specify (RegEx)
* pattern: command lines used to start the processes (RegEx)
* Metrics collected by procstat plugin begins with “procstat” prefix (e.g., procstat_cpu_time, procstat_cpu_usage, …)
40
Status Checks
* Automated checks to identify hardware and software issues
41
System Status Checks
Monitors Problems with AWS Systems (Software/hardware issues on the physical host, loss of power,....)
Check Personal Health Dashboard for any schedule. critical maintenance by AWS to your instance's host
Resolution: stop and start the instance (instance migrated to a new host)
42
Instance Status Checks
Monitors software/network configuration of your instance (invalid network configuration, exhausted memory,...)
43
Attached EBS Status Checks
Monitors EBS Volumes attached to your instance (reachable & complete I/O operations)
Resolution: Reboot the instance or replace affected EBS Volumes.
44
Status Checks - CW Metrics & Recovery
CloudWatch Metrics (1 minute interval)
* StatusCheckFailed_System
* StatusCheckFailed_Instance
EC2 Instance
* StatusCheckFailed_AttachedEBS
* StatusCheckFailed (for any)
Option 1: CloudWatch Alarm
Recover EC2 instance with the same private/public IP, EIP, metadata, and Placement Group
* Send notifications using SNS
* Option 2: Auto Scaling Group
* Set min/max/desired 1 to recover an instance but won't keep the same private and elastic IP.
45
Why EC2 Hibernate?
* We know we can stop, terminate instances
* Stop – s the data on disk (EBS) is kept intact in the next start
* Terminate – any EBS volumes (root) also set-up to be destroyed is lost
* On start, the following happens:
* First start: the OS boots & the EC2 User Data script is run
* Following starts: the OS boots up
* Then your application starts, caches get warmed up, and that can take time!
46
EC2 Hibernate - Behavior
Introducing EC2 Hibernate:
* The in-memory (RAM) state is preserved
* The instance boot is much faster!
(the OS is not stopped / restarted)
* Under the hood: the RAM state is written
to a file in the root EBS volume
* The root EBS volume must be encrypted
Use cases:
* Long-running processing
* Saving the RAM state
* Services that take time to initialize
47
EC2 Hibernate – Good to know
* Supported Instance Families – C3, C4, C5, I3, M3, M4, R3, R4, T2, T3, …
* Instance RAM Size – must be less than 150 GB.
* Instance Size – not supported for bare metal instances.
* AMI – Amazon Linux 2, Linux AMI, Ubuntu, RHEL, CentOS & Windows…
* Root Volume – must be EBS, encrypted, not instance store, and large
* Available for On-Demand, Reserved and Spot Instances
* An instance can NOT be hibernated more than 60 days
AWS SysOps
flashcards
Decks in class (2)
# Cards
