Amazon CloudWatch – Metrics, Alarms & Dashboards

December 31, 2022 ~ Last updated on : June 26, 2026 ~ jayendrapatil ~ 51 Comments

CloudWatch namespaces are containers for metrics.
Metrics in different namespaces are isolated from each other, so that metrics from different applications are not mistakenly aggregated into the same statistics.

AWS namespaces all follow the convention AWS/<service>, for e.g. AWS/EC2 and AWS/ELB
Namespace names must be fewer than 256 characters in length.

There is no default namespace. Each data element put into CloudWatch must specify a namespace.

Metrics

Metric is the fundamental concept in CloudWatch.
Uniquely defined by a name, a namespace, and one or more dimensions.
Represents a time-ordered set of data points published to CloudWatch.

Each data point has a time stamp, and (optionally) a unit of measure.
Data points can be either custom metrics or metrics from other services in AWS.
Statistics can be retrieved about those data points as an ordered set of time-series data that occur within a specified time window.

When the statistics are requested, the returned data stream is identified by namespace, metric name, dimension, and (optionally) the unit.
Metrics exist only in the region in which they are created.
CloudWatch stores the metric data for two weeks
Metrics cannot be deleted, but they automatically expire after 15 months, if no new data is published to them.

Metric retention is as follows
- Data points with a period of less than 60 seconds are available for 3 hours. These data points are high-resolution custom metrics.
- Data points with a 60 secs (1 min) period are available for 15 days
- Data points with a 300 secs (5 min) period are available for 63 days
- Data points with a 3600 secs (1 hour) period are available for 455 days (15 months)

Dimensions

A dimension is a name/value pair that uniquely identifies a metric.

Every metric has specific characteristics that describe it, and you can think of dimensions as categories for those characteristics.
Dimensions help design a structure for the statistics plan.
Dimensions are part of the unique identifier for a metric, whenever a unique name pair is added to one of the metrics, a new metric is created.
Dimensions can be used to filter result sets that CloudWatch query returns.

A metric can be assigned up to ten dimensions to a metric.

Time Stamps

Each metric data point must be marked with a time stamp to identify the data point on a time series.
Timestamp can be up to two weeks in the past and up to two hours into the future.

If no timestamp is provided, a time stamp based on the time the data element was received is created.
All times reflect the UTC time zone when statistics are retrieved

Resolution

Each metric is one of the following:
- Standard resolution, with data having a one-minute granularity
- High resolution, with data at a granularity of one second

Units

Units represent the statistic’s unit of measure e.g. count, bytes, %, etc

Statistics

Statistics are metric data aggregations over specified periods of time

Aggregations are made using the namespace, metric name, dimensions, and the data point unit of measure, within the specified time period

Periods

Period is the length of time associated with a specific statistic.
Each statistic represents an aggregation of the metrics data collected for a specified period of time.

Although periods are expressed in seconds, the minimum granularity for a period is one minute.

Aggregation

CloudWatch aggregates statistics according to the period length specified in calls to GetMetricStatistics.
Multiple data points can be published with the same or similar time stamps. CloudWatch aggregates them by period length when the statistics about those data points are requested.

Aggregated statistics are only available when using detailed monitoring.
Instances that use basic monitoring are not included in the aggregates
CloudWatch does not aggregate data across regions.

Alarms

Alarms can automatically initiate actions on behalf of the user, based on specified parameters.
Alarm watches a single metric over a specified time period, and performs one or more actions based on the value of the metric relative to a given threshold over a number of time periods.

Alarms invoke actions for sustained state changes only i.e. the state must have changed and been maintained for a specified number of periods.
Action can be a
- SNS notification
- Auto Scaling policies
- EC2 action – stop or terminate EC2 instances
- Lambda function invocation (same or cross-account)
- Systems Manager OpsItems or Incident Manager incidents
After an alarm invokes an action due to a change in state, its subsequent behavior depends on the type of action associated with the alarm.
- For Auto Scaling policy notifications, the alarm continues to invoke the action for every period that the alarm remains in the new state.
- For SNS notifications, no additional actions are invoked.
An alarm has three possible states:
- OK—The metric is within the defined threshold
- ALARM—The metric is outside of the defined threshold
- INSUFFICIENT_DATA—Alarm has just started, the metric is not available, or not enough data is available for the metric to determine the alarm state

Alarms exist only in the region in which they are created.
Alarm actions must reside in the same region as the alarm
Alarm history is available for the last 14 days.
Alarm can be tested by setting it to any state using the SetAlarmState API (mon-set-alarm-state command). This temporary state change lasts only until the next alarm comparison occurs.

Alarms can be disabled and enabled using the DisableAlarmActions and EnableAlarmActions APIs (mon-disable-alarm-actions and mon-enable-alarm-actions commands).
CloudWatch now allows alarming on data up to 7 days old (increased from previous 24-hour limit), with a period of at least 3,600 seconds (1 hour).

Composite Alarms

Composite alarms determine their state by monitoring the states of other alarms.

Rules are defined to combine the status of monitored alarms using Boolean logic (AND, OR, NOT).
Composite alarms help reduce alarm noise by taking actions only at an aggregated level.
A composite alarm’s rule expression can include as many as 100 underlying alarms.

Any single alarm can be included in the rule expressions of as many as 150 composite alarms.
Composite alarms can:
- Send Amazon SNS notifications when they change state
- Create Systems Manager OpsItems or incidents when they go into ALARM state
- Start CloudWatch Investigations for root cause analysis

Composite alarms cannot perform EC2 actions or Auto Scaling actions.

Anomaly Detection

CloudWatch anomaly detection analyzes past metric data and creates a model of expected values.
Alarms can be created based on anomaly detection to alert when metric values are outside the expected band.

The model adjusts for time-of-day patterns, day-of-week patterns, and trends over time.
Useful for metrics without a fixed threshold, such as CPU utilization with seasonal patterns.

Regions

CloudWatch does not aggregate data across regions. Therefore, metrics are completely separate between regions.

CloudWatch Metrics Insights

CloudWatch Metrics Insights is a high-performance SQL-based query engine to query metrics at scale.
Enables identification of trends and patterns across all CloudWatch metrics in real time.
Provides access to up to two weeks of historical data for trend analysis.

Supports functions including AVG, COUNT, MAX, MIN, and SUM.
Can be combined with CloudWatch Alarms to set up dynamic alarms that consistently monitor fast-moving environments.
Supports natural language query generation powered by generative AI – users can ask questions in plain English to generate metric queries.

Custom Metrics

CloudWatch allows publishing custom metrics with put-metric-data CLI command (or its Query API equivalent PutMetricData)

CloudWatch creates a new metric if put-metric-data is called with a new metric name, else it associates the data with the specified existing metric
put-metric-data command can only publish one data point per call
CloudWatch stores data about a metric as a series of data points and each data point has an associated time stamp

Creating a new metric using the put-metric-data command, can take up to two minutes before statistics can be retrieved on the new metric using the get-metric-statistics command and can take up to fifteen minutes before the new metric appears in the list of metrics retrieved using the list-metrics command.
CloudWatch allows publishing
- Single data point
  - Data points can be published with time stamps as granular as one-thousandth of a second, CloudWatch aggregates the data to a minimum granularity of one minute
  - CloudWatch records the average (sum of all items divided by number of items) of the values received for every 1-minute period, as well as number of samples, maximum value, and minimum value for the same time period
  - CloudWatch uses one-minute boundaries when aggregating data points
- Aggregated set of data points called a statistics set
  - Data can also be aggregated before being published to CloudWatch
  - Aggregating data minimizes the number of calls reducing it to a single call per minute with the statistic set of data
  - Statistics include Sum, Average, Minimum, Maximum, SampleCount
If the application produces data that is more sporadic and have periods that have no associated data, either a the value zero (0) or no value at all can be published

However, it can be helpful to publish zero instead of no value
- to monitor the health of your application for e.g. alarm can be configured to notify if no metrics published every 5 minutes
- to track the total number of data points
- to have statistics such as minimum and average to include data points with the value 0.

CloudWatch Dashboards

CloudWatch dashboards are customizable home pages in the CloudWatch console used to monitor the resources in a single view, even those resources that are spread across different Regions.
Dashboards can be used to create customized views of the metrics and alarms for the AWS resources.
Dashboards can help to create
- A single view for selected metrics and alarms to help assess the health of the resources and applications across one or more Regions.
- An operational playbook that provides guidance for team members during operational events about how to respond to specific incidents.
- A common view of critical resource and application measurements that can be shared by team members for faster communication flow during operational events.

CloudWatch Cross-Account Observability

CloudWatch cross-account observability helps monitor and troubleshoot applications that span multiple accounts within a Region.
Uses Observability Access Manager (OAM) to create and manage links between source and monitoring accounts.
Cross-account observability includes monitoring and source accounts
- A monitoring account is a central AWS account that can view and interact with observability data generated from source accounts.
- A source account is an individual AWS account that generates observability data for the resources that reside in it.
- Source accounts share their observability data with the monitoring account which can include the following types of telemetry:
  - Metrics in CloudWatch
  - Log groups in CloudWatch Logs
  - Traces in AWS X-Ray
  - Application Signals services and SLOs

Available in all commercial AWS Regions and AWS GovCloud (US) Regions at no extra cost for logs and metrics.
Cross-Account and Cross-Region Log Centralization (2025) allows copying log data from multiple accounts and regions into a single destination account for consolidated analysis.

CloudWatch Agent

CloudWatch Agent helps collect metrics and logs from EC2 instances and on-premises servers and push them to CloudWatch.

Logs collected by the unified agent are processed and stored in CloudWatch Logs.
Supports collecting metrics and traces via OpenTelemetry Protocol (OTLP).
Can collect from AWS, on-premises, or multicloud environments.

CloudWatch Logs

Refer blog post @ CloudWatch Logs

CloudWatch Application Signals

CloudWatch Application Signals (GA June 2024) provides application performance monitoring (APM) capabilities.
Automatically instruments applications on AWS to track application performance against business objectives.

Provides pre-built, standardized dashboards showing key metrics: volume, availability, latency, faults, and errors.
Supports Service Level Objectives (SLOs) to define and track application performance targets.
Application map automatically discovers and visualizes application topology without requiring instrumentation.

Supports runtime metrics for Java, Python, and .NET applications.
Supports serverless applications on AWS Lambda.
Works with cross-account observability for centralized monitoring.

CloudWatch Internet Monitor

Internet Monitor provides near-continuous internet measurements for internet traffic tailored to your workload footprint on AWS.
Provides availability and performance metrics (performance score, availability score, round-trip time).
Alerts when internet issues affect application users.
Helps troubleshoot and mitigate network problems before they impact customers.

Creates aggregated CloudWatch metrics for traffic to your application, by AWS Region and edge location.

CloudWatch Investigations

CloudWatch Investigations (GA June 2025) is an AI-powered root cause analysis feature.
Powered by Amazon Q Developer, it analyzes metrics, logs, traces, and deployment events.

Helps find root cause without manually querying multiple sources of telemetry.
Identifies anomalies, surfaces related signals, suggests remediation steps.
Can be triggered automatically from a CloudWatch alarm or created manually.
Supports integration with Slack for DevOps troubleshooting workflows.
Significantly reduces Mean Time to Resolution (MTTR).

CloudWatch Pipelines

CloudWatch Pipelines is a fully managed service that ingests, transforms, and routes log data to CloudWatch without managing infrastructure.

Supports log data from AWS services, third-party applications, and custom sources.
Provides processors for parsing, transforming, and enriching log data.
Supports conditional processing and drop event processing for fine-grained control.
Includes compliance and governance capabilities for data integrity and access control.

Supports AI-assisted configuration of processors via natural language descriptions.
Supports OCSF (Open Cybersecurity Schema Framework) and OpenTelemetry formats.

CloudWatch OpenTelemetry Support

CloudWatch now supports native OpenTelemetry (OTel) metrics ingestion (preview, April 2026).
Enables sending metrics directly using OTLP without custom conversion logic or additional tooling.
Supports high-cardinality metrics with up to 150 labels per metric.

Supports metric types: gauge, sum, histogram, and exponential histogram.
Metrics can be queried using PromQL (Prometheus Query Language).
Automatic AWS vended metric enrichment adds AWS context to ingested metrics.
Completes OTel support across all three observability pillars (traces, logs, metrics).

CloudWatch Database Insights

CloudWatch Database Insights (launched December 2024) provides comprehensive database observability from fleet-level to instance-level.

Supports cross-account and cross-region database fleet monitoring.
Offers two service tiers:
- Standard mode – 7 days of data retention at no additional cost
- Advanced mode – Extended 15-month retention with enhanced capabilities
Transitioning from RDS Performance Insights to Database Insights.

CloudWatch Evidently (Deprecated)

⚠️ CloudWatch Evidently reached End of Life on October 17, 2025.

Alternative: Use AWS Systems Manager AppConfig for feature flag management.

CloudWatch Events → Amazon EventBridge

CloudWatch Events has been superseded by Amazon EventBridge.
EventBridge is the evolution of CloudWatch Events with additional capabilities including partner events, Schema Registry, and EventBridge Pipes.
New features are added to EventBridge only and not to CloudWatch Events.

The underlying API and service remain the same – changes in either console appear in both.
AWS recommends using EventBridge for all new event-driven architectures.

CloudWatch Supported Services

Refer blog post @ CloudWatch Supported Services

Accessing CloudWatch

CloudWatch can be accessed using
- AWS CloudWatch console
- CloudWatch CLI
- AWS CLI
- CloudWatch API
- AWS SDKs
- AWS Console Mobile App (supports interactive alarm graphs, AI log summaries, and natural language log search)

CloudWatch Key Updates (2024-2026)

Unified Data Management (Dec 2025) – CloudWatch provides unified management and analytics for operational, security, and compliance data across AWS and third-party sources.
Organization-wide EC2 Detailed Monitoring (Mar 2026) – Enablement rules automatically enable detailed monitoring for existing and new EC2 instances matching the rule scope.
Natural Language Query Generation (GA June 2024) – Powered by generative AI for both Logs Insights and Metrics Insights, supporting CloudWatch Logs Insights QL, OpenSearch PPL, and OpenSearch SQL.

Integrated Analytics with OpenSearch (Dec 2024) – Supports SQL and PPL query languages for log analytics with JOIN, sub-queries, and SQL functions.
Cross-Region Telemetry Auditing (Apr 2026) – Enable telemetry auditing for account/organization across all regions with enablement rules.
Lambda Tiered Pricing for Logs (May 2025) – Reduces Lambda CloudWatch logging costs with tiered pricing and additional logging destinations.

AWS Certification Exam Practice Questions

Questions are collected from Internet and the answers are marked as per my knowledge and understanding (which might differ with yours).

AWS services are updated everyday and both the answers and questions might be outdated soon, so research accordingly.

AWS exam questions are not updated to keep up the pace with AWS updates, so even if the underlying feature has changed the question might not be updated

Open to further feedback, discussion and correction.

Questions are collected from Internet and the answers are marked as per my knowledge and understanding (which might differ with yours).
AWS services are updated everyday and both the answers and questions might be outdated soon, so research accordingly.
AWS exam questions are not updated to keep up the pace with AWS updates, so even if the underlying feature has changed the question might not be updated

Open to further feedback, discussion and correction.

A company needs to monitor the read and write IOPs metrics for their AWS MySQL RDS instance and send real-time alerts to their operations team. Which AWS services can accomplish this? Choose 2 answers
1. Amazon Simple Email Service (Cannot be integrated with CloudWatch directly)
2. Amazon CloudWatch
3. Amazon Simple Queue Service
4. Amazon Route 53
5. Amazon Simple Notification Service
A customer needs to capture all client connection information from their load balancer every five minutes. The company wants to use this data for analyzing traffic patterns and troubleshooting their applications. Which of the following options meets the customer requirements?
1. Enable AWS CloudTrail for the load balancer.
2. Enable access logs on the load balancer. (Refer link)
3. Install the Amazon CloudWatch Logs agent on the load balancer.
4. Enable Amazon CloudWatch metrics on the load balancer (does not provide Client connection information)
A user is running a batch process on EBS backed EC2 instances. The batch process starts a few instances to process Hadoop Map reduce jobs, which can run between 50 – 600 minutes or sometimes for more time. The user wants to configure that the instance gets terminated only when the process is completed. How can the user configure this with CloudWatch?
1. Setup the CloudWatch action to terminate the instance when the CPU utilization is less than 5%
2. Setup the CloudWatch with Auto Scaling to terminate all the instances
3. Setup a job which terminates all instances after 600 minutes
4. It is not possible to terminate instances automatically

A user has two EC2 instances running in two separate regions. The user is running an internal memory management tool, which captures the data and sends it to CloudWatch in US East, using a CLI with the same namespace and metric. Which of the below mentioned options is true with respect to the above statement?
1. The setup will not work as CloudWatch cannot receive data across regions
2. CloudWatch will receive and aggregate the data based on the namespace and metric
3. CloudWatch will give an error since the data will conflict due to two sources
4. CloudWatch will take the data of the server, which sends the data first
A user is sending the data to CloudWatch using the CloudWatch API. The user is sending data 90 minutes in the future. What will CloudWatch do in this case?
1. CloudWatch will accept the data
2. It is not possible to send data of the future
3. It is not possible to send the data manually to CloudWatch
4. The user cannot send data for more than 60 minutes in the future
A user is having data generated randomly based on a certain event. The user wants to upload that data to CloudWatch. It may happen that event may not have data generated for some period due to randomness. Which of the below mentioned options is a recommended option for this case?
1. For the period when there is no data, the user should not send the data at all
2. For the period when there is no data the user should send a blank value
3. For the period when there is no data the user should send the value as 0 (Refer User Guide)
4. The user must upload the data to CloudWatch as having no data for some period will cause an error at CloudWatch monitoring
A user has a weighing plant. The user measures the weight of some goods every 5 minutes and sends data to AWS CloudWatch for monitoring and tracking. Which of the below mentioned parameters is mandatory for the user to include in the request list?
1. Value
2. Namespace (refer put-metric request)
3. Metric Name
4. Timezone
A user has a refrigerator plant. The user is measuring the temperature of the plant every 15 minutes. If the user wants to send the data to CloudWatch to view the data visually, which of the below mentioned statements is true with respect to the information given above?
1. The user needs to use AWS CLI or API to upload the data
2. The user can use the AWS Import Export facility to import data to CloudWatch
3. The user will upload data from the AWS console
4. The user cannot upload data to CloudWatch since it is not an AWS service metric
A user has launched an EC2 instance. The user is planning to setup the CloudWatch alarm. Which of the below mentioned actions is not supported by the CloudWatch alarm?
1. Notify the Auto Scaling launch config to scale up
2. Send an SMS using SNS
3. Notify the Auto Scaling group to scale down
4. Stop the EC2 instance
A user has a refrigerator plant. The user is measuring the temperature of the plant every 15 minutes. If the user wants to send the data to CloudWatch to view the data visually, which of the below mentioned statements is true with respect to the information given above?
1. The user needs to use AWS CLI or API to upload the data
2. The user can use the AWS Import Export facility to import data to CloudWatch
3. The user will upload data from the AWS console
4. The user cannot upload data to CloudWatch since it is not an AWS service metric
A user is trying to aggregate all the CloudWatch metric data of the last 1 week. Which of the below mentioned statistics is not available for the user as a part of data aggregation?
1. Aggregate
2. Sum
3. Sample data
4. Average
A user has setup a CloudWatch alarm on an EC2 action when the CPU utilization is above 75%. The alarm sends a notification to SNS on the alarm state. If the user wants to simulate the alarm action how can he achieve this?
1. Run activities on the CPU such that its utilization reaches above 75%
2. From the AWS console change the state to ‘Alarm’
3. The user can set the alarm state to ‘Alarm’ using CLI
4. Run the SNS action manually
A user is publishing custom metrics to CloudWatch. Which of the below mentioned statements will help the user understand the functionality better?
1. The user can use the CloudWatch Import tool
2. The user should be able to see the data in the console after around 15 minutes
3. If the user is uploading the custom data, the user must supply the namespace, timezone, and metric name as part of the command
4. The user can view as well as upload data using the console, CLI and APIs
An application that you are managing has EC2 instances and DynamoDB tables deployed to several AWS Regions. In order to monitor the performance of the application globally, you would like to see two graphs 1) Avg CPU Utilization across all EC2 instances and 2) Number of Throttled Requests for all DynamoDB tables. How can you accomplish this? [PROFESSIONAL]
1. Tag your resources with the application name, and select the tag name as the dimension in the CloudWatch Management console to view the respective graphs (CloudWatch metrics are regional)
2. Use the CloudWatch CLI tools to pull the respective metrics from each regional endpoint. Aggregate the data offline & store it for graphing in CloudWatch. (Note: CloudWatch cross-account cross-region dashboards and log centralization are now available but this question tests the concept that metrics are regional)
3. Add SNMP traps to each instance and DynamoDB table. Leverage a central monitoring server to capture data from each instance and table. Put the aggregate data into CloudWatch for graphing (Can’t add SNMP traps to DynamoDB as it is a managed service)
4. Add a CloudWatch agent to each instance and attach one to each DynamoDB table. When configuring the agent set the appropriate application name & view the graphs in CloudWatch. (Can’t add agents to DynamoDB as it is a managed service)
You have set up Individual AWS accounts for each project. You have been asked to make sure your AWS Infrastructure costs do not exceed the budget set per project for each month. Which of the following approaches can help ensure that you do not exceed the budget each month? [PROFESSIONAL]
1. Consolidate your accounts so you have a single bill for all accounts and projects (Consolidation will not help limit per account)
2. Set up auto scaling with CloudWatch alarms using SNS to notify you when you are running too many Instances in a given account (many instances do not directly map to cost and would not give exact cost)
3. Set up CloudWatch billing alerts for all AWS resources used by each project, with a notification occurring when the amount for each resource tagged to a particular project matches the budget allocated to the project. (as each project already has a account, no need for resource tagging)
4. Set up CloudWatch billing alerts for all AWS resources used by each account, with email notifications when it hits 50%. 80% and 90% of its budgeted monthly spend
You meet once per month with your operations team to review the past month’s data. During the meeting, you realize that 3 weeks ago, your monitoring system which pings over HTTP from outside AWS recorded a large spike in latency on your 3-tier web service API. You use DynamoDB for the database layer, ELB, EBS, and EC2 for the business logic tier, and SQS, ELB, and EC2 for the presentation layer. Which of the following techniques will NOT help you figure out what happened?
1. Check your CloudTrail log history around the spike’s time for any API calls that caused slowness.
2. Review CloudWatch Metrics graphs to determine which component(s) slowed the system down. (Note: CloudWatch metric data is retained for 15 months. Data at 1-min resolution is available for 15 days, 5-min for 63 days, and 1-hour for 455 days. The 3-week old data would still be available at 5-min or 1-hour resolution.)
3. Review your ELB access logs in S3 to see if any ELBs in your system saw the latency.
4. Analyze your logs to detect bursts in traffic at that time.
You have a high security requirement for your AWS accounts. What is the most rapid and sophisticated setup you can use to react to AWS API calls to your account?
1. Subscription to AWS Config via an SNS Topic. Use a Lambda Function to perform in-flight analysis and reactivity to changes as they occur.
2. Global AWS CloudTrail setup delivering to S3 with an SNS subscription to the deliver notifications, pushing into a Lambda, which inserts records into an ELK stack for analysis.
3. Use a CloudWatch Rule ScheduleExpression to periodically analyze IAM credential logs. Push the deltas for events into an ELK stack and perform ad-hoc analysis there.
4. Amazon EventBridge Rules (formerly CloudWatch Events Rules), which trigger based on all AWS API calls, submitting all events to an AWS Kinesis Stream for arbitrary downstream analysis. (EventBridge (evolution of CloudWatch Events) allows subscription to AWS API calls, and direction of these events into Kinesis Streams. This allows a unified, near real-time stream for all API calls. Refer link)
To monitor API calls against our AWS account by different users and entities, we can use ____ to create a history of calls in bulk for later review, and use ____ for reacting to AWS API calls in real-time.
1. AWS Config; AWS Inspector
2. AWS CloudTrail; AWS Config
3. AWS CloudTrail; Amazon EventBridge (formerly CloudWatch Events) (CloudTrail is a batch API call collection service, EventBridge enables real-time monitoring of calls through the Rules object interface. Refer link)
4. AWS Config; AWS Lambda
You are hired as the new head of operations for a SaaS company. Your CTO has asked you to make debugging any part of your entire operation simpler and as fast as possible. She complains that she has no idea what is going on in the complex, service-oriented architecture, because the developers just log to disk, and it’s very hard to find errors in logs on so many services. How can you best meet this requirement and satisfy your CTO? [PROFESSIONAL]
1. Copy all log files into AWS S3 using a cron job on each instance. Use an S3 Notification Configuration on the <code>PutBucket</code> event and publish events to AWS Lambda. Use the Lambda to analyze logs as soon as they come in and flag issues. (is not fast in search and introduces delay)
2. Begin using CloudWatch Logs on every service. Stream all Log Groups into S3 objects. Use AWS EMR cluster jobs to perform adhoc MapReduce analysis and write new queries when needed. (is not fast in search and introduces delay)
3. Copy all log files into AWS S3 using a cron job on each instance. Use an S3 Notification Configuration on the <code>PutBucket</code> event and publish events to AWS Kinesis. Use Apache Spark on AWS EMR to perform at-scale stream processing queries on the log chunks and flag issues. (is not fast in search and introduces delay)
4. Begin using CloudWatch Logs on every service. Stream all Log Groups into an Amazon OpenSearch Service Domain running Kibana and perform log analysis on a search cluster. (OpenSearch (formerly Elasticsearch) with Kibana/OpenSearch Dashboards is designed for real-time, ad-hoc log analysis and aggregation. Note: CloudWatch now also offers integrated analytics with OpenSearch Service directly.)
Your EC2-Based Multi-tier application includes a monitoring instance that periodically makes application-level read only requests of various application components and if any of those fail more than three times 30 seconds calls CloudWatch to fire an alarm, and the alarm notifies your operations team by email and SMS of a possible application health problem. However, you also need to watch the watcher -the monitoring instance itself – and be notified if it becomes unhealthy. Which of the following is a simple way to achieve that goal? [PROFESSIONAL]
1. Run another monitoring instance that pings the monitoring instance and fires a CloudWatch alarm that notifies your operations team should the primary monitoring instance become unhealthy.
2. Set a CloudWatch alarm based on EC2 system and instance status checks and have the alarm notify your operations team of any detected problem with the monitoring instance.
3. Set a CloudWatch alarm based on the CPU utilization of the monitoring instance and have the alarm notify your operations team if the CPU usage exceeds 50% for more than one minute: then have your monitoring application go into a CPU-bound loop should it detect any application problems.
4. Have the monitoring instances post messages to an SQS queue and then dequeue those messages on another instance should the queue cease to have new messages, the second instance should first terminate the original monitoring instance start another backup monitoring instance and assume the role of the previous monitoring instance and beginning adding messages to the SQS queue.
A company wants to reduce alarm noise in their monitoring system. They have multiple alarms for different components of a web application. Which CloudWatch feature should they use to consolidate multiple alarms into a single notification?
1. CloudWatch Metrics Insights
2. CloudWatch Composite Alarms (Composite alarms combine multiple alarm states using Boolean logic to reduce noise and send notifications only at an aggregated level)
3. CloudWatch Anomaly Detection
4. CloudWatch Dashboards
A DevOps team wants to automatically identify root causes when CloudWatch alarms fire, correlating metrics, logs, and deployment events without manual investigation. Which CloudWatch feature should they use?
1. CloudWatch Logs Insights
2. CloudWatch Contributor Insights
3. CloudWatch Investigations (Powered by Amazon Q Developer, Investigations uses AI to analyze multiple telemetry sources and suggest root causes and remediation steps)
4. CloudWatch Application Signals
An organization wants to send application metrics to CloudWatch using open standards without vendor-specific conversion logic. They need support for high-cardinality metrics with many labels. Which approach should they use?
1. Use CloudWatch PutMetricData API with custom metrics
2. Use CloudWatch Metric Streams to export to a third-party tool
3. Send metrics directly using the OpenTelemetry Protocol (OTLP) to CloudWatch’s native OTel endpoint (CloudWatch supports native OTLP metrics ingestion with up to 150 labels per metric, queryable via PromQL)
4. Use the CloudWatch Agent with StatsD protocol
A company wants to ingest, transform, and route log data from multiple sources to CloudWatch without managing any infrastructure. They need to parse complex log formats and enrich the data during ingestion. Which service should they use?
1. CloudWatch Logs Subscription Filters
2. Amazon Kinesis Data Firehose
3. CloudWatch Pipelines (CloudWatch Pipelines is a fully managed service that ingests, transforms, and routes log data with support for parsing, conditional processing, and AI-assisted processor configuration)
4. CloudWatch Log Transformers only
A team needs to monitor their application’s availability, latency, and error rate with pre-built dashboards and set up Service Level Objectives (SLOs) without building custom monitoring. Which CloudWatch capability provides this?
1. CloudWatch Synthetics
2. CloudWatch Container Insights
3. CloudWatch Internet Monitor
4. CloudWatch Application Signals (Application Signals provides automatic instrumentation, pre-built APM dashboards with golden metrics, and SLO tracking without custom setup)

Amazon CloudWatch

CloudWatch monitors AWS resources and applications in real time.
CloudWatch can be used to collect and track metrics, which are the variables to be measured for resources and applications.
CloudWatch is basically a metrics repository where the metrics can be inserted and statistics retrieved based on those metrics.
In addition to monitoring the built-in metrics that come with AWS, custom metrics can also be monitored
CloudWatch provides system-wide visibility into resource utilization, application performance, and operational health.
By default, CloudWatch stores the log data indefinitely, and the retention can be changed for each log group at any time.
CloudWatch alarms can be configured
- to send notifications or
- to automatically make changes to the resources based on defined rules
CloudWatch dashboards are customizable home pages in the CloudWatch console used to monitor the resources in a single view, even those resources that are spread across different Regions.
CloudWatch Agent helps collect metrics and logs from EC2 instances and on-premises servers and push them to CloudWatch.
CloudWatch now supports native OpenTelemetry (OTel) metrics, enabling metrics to be sent directly using the OpenTelemetry Protocol (OTLP) without custom conversion logic.
CloudWatch supports PromQL (Prometheus Query Language) for querying metrics including OTel-ingested and AWS vended metrics.
CloudWatch Investigations uses AI-powered root cause analysis to help identify anomalies, surface related signals, and suggest remediation steps.

CloudWatch Architecture

CloudWatch collects various metrics from various resources
These metrics, as statistics, are available to the user through Console, CLI
CloudWatch allows the creation of alarms with defined rules
- to perform actions to auto-scaling or stop, start, or terminate instances
- to send notifications using SNS actions on your behalf

CloudWatch Concepts

Namespaces

CloudWatch namespaces are containers for metrics.
Metrics in different namespaces are isolated from each other, so that metrics from different applications are not mistakenly aggregated into the same statistics.
AWS namespaces all follow the convention AWS/<service>, for e.g. AWS/EC2 and AWS/ELB
Namespace names must be fewer than 256 characters in length.
There is no default namespace. Each data element put into CloudWatch must specify a namespace.

Metrics

Metric is the fundamental concept in CloudWatch.
Uniquely defined by a name, a namespace, and one or more dimensions.
Represents a time-ordered set of data points published to CloudWatch.
Each data point has a time stamp, and (optionally) a unit of measure.
Data points can be either custom metrics or metrics from other services in AWS.
Statistics can be retrieved about those data points as an ordered set of time-series data that occur within a specified time window.
When the statistics are requested, the returned data stream is identified by namespace, metric name, dimension, and (optionally) the unit.
Metrics exist only in the region in which they are created.
CloudWatch stores the metric data for two weeks
Metrics cannot be deleted, but they automatically expire after 15 months, if no new data is published to them.
Metric retention is as follows
- Data points with a period of less than 60 seconds are available for 3 hours. These data points are high-resolution custom metrics.
- Data points with a 60 secs (1 min) period are available for 15 days
- Data points with a 300 secs (5 min) period are available for 63 days
- Data points with a 3600 secs (1 hour) period are available for 455 days (15 months)

Dimensions

A dimension is a name/value pair that uniquely identifies a metric.
Every metric has specific characteristics that describe it, and you can think of dimensions as categories for those characteristics.
Dimensions help design a structure for the statistics plan.
Dimensions are part of the unique identifier for a metric, whenever a unique name pair is added to one of the metrics, a new metric is created.
Dimensions can be used to filter result sets that CloudWatch query returns.
A metric can be assigned up to ten dimensions to a metric.

Time Stamps

Each metric data point must be marked with a time stamp to identify the data point on a time series.
Timestamp can be up to two weeks in the past and up to two hours into the future.
If no timestamp is provided, a time stamp based on the time the data element was received is created.
All times reflect the UTC time zone when statistics are retrieved

Resolution

Each metric is one of the following:
- Standard resolution, with data having a one-minute granularity
- High resolution, with data at a granularity of one second

Units

Units represent the statistic’s unit of measure e.g. count, bytes, %, etc

Statistics

Statistics are metric data aggregations over specified periods of time
Aggregations are made using the namespace, metric name, dimensions, and the data point unit of measure, within the specified time period

Periods

Period is the length of time associated with a specific statistic.
Each statistic represents an aggregation of the metrics data collected for a specified period of time.
Although periods are expressed in seconds, the minimum granularity for a period is one minute.

Aggregation

CloudWatch aggregates statistics according to the period length specified in calls to GetMetricStatistics.
Multiple data points can be published with the same or similar time stamps. CloudWatch aggregates them by period length when the statistics about those data points are requested.
Aggregated statistics are only available when using detailed monitoring.
Instances that use basic monitoring are not included in the aggregates
CloudWatch does not aggregate data across regions.

Alarms

Alarms can automatically initiate actions on behalf of the user, based on specified parameters.
Alarm watches a single metric over a specified time period, and performs one or more actions based on the value of the metric relative to a given threshold over a number of time periods.
Alarms invoke actions for sustained state changes only i.e. the state must have changed and been maintained for a specified number of periods.
Action can be a
- SNS notification
- Auto Scaling policies
- EC2 action – stop or terminate EC2 instances
- Lambda function invocation (same or cross-account)
- Systems Manager OpsItems or Incident Manager incidents
After an alarm invokes an action due to a change in state, its subsequent behavior depends on the type of action associated with the alarm.
- For Auto Scaling policy notifications, the alarm continues to invoke the action for every period that the alarm remains in the new state.
- For SNS notifications, no additional actions are invoked.
An alarm has three possible states:
- OK—The metric is within the defined threshold
- ALARM—The metric is outside of the defined threshold
- INSUFFICIENT_DATA—Alarm has just started, the metric is not available, or not enough data is available for the metric to determine the alarm state
Alarms exist only in the region in which they are created.
Alarm actions must reside in the same region as the alarm
Alarm history is available for the last 14 days.
Alarm can be tested by setting it to any state using the SetAlarmState API (mon-set-alarm-state command). This temporary state change lasts only until the next alarm comparison occurs.
Alarms can be disabled and enabled using the DisableAlarmActions and EnableAlarmActions APIs (mon-disable-alarm-actions and mon-enable-alarm-actions commands).
CloudWatch now allows alarming on data up to 7 days old (increased from previous 24-hour limit), with a period of at least 3,600 seconds (1 hour).

Composite Alarms

Composite alarms determine their state by monitoring the states of other alarms.
Rules are defined to combine the status of monitored alarms using Boolean logic (AND, OR, NOT).
Composite alarms help reduce alarm noise by taking actions only at an aggregated level.
A composite alarm’s rule expression can include as many as 100 underlying alarms.
Any single alarm can be included in the rule expressions of as many as 150 composite alarms.
Composite alarms can:
- Send Amazon SNS notifications when they change state
- Create Systems Manager OpsItems or incidents when they go into ALARM state
- Start CloudWatch Investigations for root cause analysis
Composite alarms cannot perform EC2 actions or Auto Scaling actions.

Anomaly Detection

CloudWatch anomaly detection analyzes past metric data and creates a model of expected values.
Alarms can be created based on anomaly detection to alert when metric values are outside the expected band.
The model adjusts for time-of-day patterns, day-of-week patterns, and trends over time.
Useful for metrics without a fixed threshold, such as CPU utilization with seasonal patterns.

Regions

CloudWatch does not aggregate data across regions. Therefore, metrics are completely separate between regions.

CloudWatch Metrics Insights

CloudWatch Metrics Insights is a high-performance SQL-based query engine to query metrics at scale.
Enables identification of trends and patterns across all CloudWatch metrics in real time.
Provides access to up to two weeks of historical data for trend analysis.
Supports functions including AVG, COUNT, MAX, MIN, and SUM.
Can be combined with CloudWatch Alarms to set up dynamic alarms that consistently monitor fast-moving environments.
Supports natural language query generation powered by generative AI – users can ask questions in plain English to generate metric queries.

Custom Metrics

CloudWatch allows publishing custom metrics with put-metric-data CLI command (or its Query API equivalent PutMetricData)
CloudWatch creates a new metric if put-metric-data is called with a new metric name, else it associates the data with the specified existing metric
put-metric-data command can only publish one data point per call
CloudWatch stores data about a metric as a series of data points and each data point has an associated time stamp
Creating a new metric using the put-metric-data command, can take up to two minutes before statistics can be retrieved on the new metric using the get-metric-statistics command and can take up to fifteen minutes before the new metric appears in the list of metrics retrieved using the list-metrics command.
CloudWatch allows publishing
- Single data point
  - Data points can be published with time stamps as granular as one-thousandth of a second, CloudWatch aggregates the data to a minimum granularity of one minute
  - CloudWatch records the average (sum of all items divided by number of items) of the values received for every 1-minute period, as well as number of samples, maximum value, and minimum value for the same time period
  - CloudWatch uses one-minute boundaries when aggregating data points
- Aggregated set of data points called a statistics set
  - Data can also be aggregated before being published to CloudWatch
  - Aggregating data minimizes the number of calls reducing it to a single call per minute with the statistic set of data
  - Statistics include Sum, Average, Minimum, Maximum, SampleCount
If the application produces data that is more sporadic and have periods that have no associated data, either a the value zero (0) or no value at all can be published
However, it can be helpful to publish zero instead of no value
- to monitor the health of your application for e.g. alarm can be configured to notify if no metrics published every 5 minutes
- to track the total number of data points
- to have statistics such as minimum and average to include data points with the value 0.

CloudWatch Dashboards

CloudWatch dashboards are customizable home pages in the CloudWatch console used to monitor the resources in a single view, even those resources that are spread across different Regions.
Dashboards can be used to create customized views of the metrics and alarms for the AWS resources.
Dashboards can help to create
- A single view for selected metrics and alarms to help assess the health of the resources and applications across one or more Regions.
- An operational playbook that provides guidance for team members during operational events about how to respond to specific incidents.
- A common view of critical resource and application measurements that can be shared by team members for faster communication flow during operational events.

CloudWatch Cross-Account Observability

CloudWatch cross-account observability helps monitor and troubleshoot applications that span multiple accounts within a Region.
Uses Observability Access Manager (OAM) to create and manage links between source and monitoring accounts.
Cross-account observability includes monitoring and source accounts
- A monitoring account is a central AWS account that can view and interact with observability data generated from source accounts.
- A source account is an individual AWS account that generates observability data for the resources that reside in it.
- Source accounts share their observability data with the monitoring account which can include the following types of telemetry:
  - Metrics in CloudWatch
  - Log groups in CloudWatch Logs
  - Traces in AWS X-Ray
  - Application Signals services and SLOs
Available in all commercial AWS Regions and AWS GovCloud (US) Regions at no extra cost for logs and metrics.
Cross-Account and Cross-Region Log Centralization (2025) allows copying log data from multiple accounts and regions into a single destination account for consolidated analysis.

CloudWatch Agent

CloudWatch Agent helps collect metrics and logs from EC2 instances and on-premises servers and push them to CloudWatch.
Logs collected by the unified agent are processed and stored in CloudWatch Logs.
Supports collecting metrics and traces via OpenTelemetry Protocol (OTLP).
Can collect from AWS, on-premises, or multicloud environments.

CloudWatch Logs

Refer blog post @ CloudWatch Logs

CloudWatch Application Signals

CloudWatch Application Signals (GA June 2024) provides application performance monitoring (APM) capabilities.
Automatically instruments applications on AWS to track application performance against business objectives.
Provides pre-built, standardized dashboards showing key metrics: volume, availability, latency, faults, and errors.
Supports Service Level Objectives (SLOs) to define and track application performance targets.
Application map automatically discovers and visualizes application topology without requiring instrumentation.
Supports runtime metrics for Java, Python, and .NET applications.
Supports serverless applications on AWS Lambda.
Works with cross-account observability for centralized monitoring.

CloudWatch Internet Monitor

Internet Monitor provides near-continuous internet measurements for internet traffic tailored to your workload footprint on AWS.
Provides availability and performance metrics (performance score, availability score, round-trip time).
Alerts when internet issues affect application users.
Helps troubleshoot and mitigate network problems before they impact customers.
Creates aggregated CloudWatch metrics for traffic to your application, by AWS Region and edge location.

CloudWatch Investigations

CloudWatch Investigations (GA June 2025) is an AI-powered root cause analysis feature.
Powered by Amazon Q Developer, it analyzes metrics, logs, traces, and deployment events.
Helps find root cause without manually querying multiple sources of telemetry.
Identifies anomalies, surfaces related signals, suggests remediation steps.
Can be triggered automatically from a CloudWatch alarm or created manually.
Supports integration with Slack for DevOps troubleshooting workflows.
Significantly reduces Mean Time to Resolution (MTTR).

CloudWatch Pipelines

CloudWatch Pipelines is a fully managed service that ingests, transforms, and routes log data to CloudWatch without managing infrastructure.
Supports log data from AWS services, third-party applications, and custom sources.
Provides processors for parsing, transforming, and enriching log data.
Supports conditional processing and drop event processing for fine-grained control.
Includes compliance and governance capabilities for data integrity and access control.
Supports AI-assisted configuration of processors via natural language descriptions.
Supports OCSF (Open Cybersecurity Schema Framework) and OpenTelemetry formats.

CloudWatch OpenTelemetry Support

CloudWatch now supports native OpenTelemetry (OTel) metrics ingestion (preview, April 2026).
Enables sending metrics directly using OTLP without custom conversion logic or additional tooling.
Supports high-cardinality metrics with up to 150 labels per metric.
Supports metric types: gauge, sum, histogram, and exponential histogram.
Metrics can be queried using PromQL (Prometheus Query Language).
Automatic AWS vended metric enrichment adds AWS context to ingested metrics.
Completes OTel support across all three observability pillars (traces, logs, metrics).

CloudWatch Database Insights

CloudWatch Database Insights (launched December 2024) provides comprehensive database observability from fleet-level to instance-level.
Supports cross-account and cross-region database fleet monitoring.
Offers two service tiers:
- Standard mode – 7 days of data retention at no additional cost
- Advanced mode – Extended 15-month retention with enhanced capabilities
Transitioning from RDS Performance Insights to Database Insights.

CloudWatch Evidently (Deprecated)

⚠️ CloudWatch Evidently reached End of Life on October 17, 2025.

Alternative: Use AWS Systems Manager AppConfig for feature flag management.

CloudWatch Events → Amazon EventBridge

CloudWatch Events has been superseded by Amazon EventBridge.
EventBridge is the evolution of CloudWatch Events with additional capabilities including partner events, Schema Registry, and EventBridge Pipes.
New features are added to EventBridge only and not to CloudWatch Events.
The underlying API and service remain the same – changes in either console appear in both.
AWS recommends using EventBridge for all new event-driven architectures.

CloudWatch Supported Services

Refer blog post @ CloudWatch Supported Services

Accessing CloudWatch

CloudWatch can be accessed using
- AWS CloudWatch console
- CloudWatch CLI
- AWS CLI
- CloudWatch API
- AWS SDKs
- AWS Console Mobile App (supports interactive alarm graphs, AI log summaries, and natural language log search)

CloudWatch Key Updates (2024-2026)

Unified Data Management (Dec 2025) – CloudWatch provides unified management and analytics for operational, security, and compliance data across AWS and third-party sources.
Organization-wide EC2 Detailed Monitoring (Mar 2026) – Enablement rules automatically enable detailed monitoring for existing and new EC2 instances matching the rule scope.
Natural Language Query Generation (GA June 2024) – Powered by generative AI for both Logs Insights and Metrics Insights, supporting CloudWatch Logs Insights QL, OpenSearch PPL, and OpenSearch SQL.
Integrated Analytics with OpenSearch (Dec 2024) – Supports SQL and PPL query languages for log analytics with JOIN, sub-queries, and SQL functions.
Cross-Region Telemetry Auditing (Apr 2026) – Enable telemetry auditing for account/organization across all regions with enablement rules.
Lambda Tiered Pricing for Logs (May 2025) – Reduces Lambda CloudWatch logging costs with tiered pricing and additional logging destinations.

AWS Certification Exam Practice Questions

Questions are collected from Internet and the answers are marked as per my knowledge and understanding (which might differ with yours).

AWS services are updated everyday and both the answers and questions might be outdated soon, so research accordingly.

AWS exam questions are not updated to keep up the pace with AWS updates, so even if the underlying feature has changed the question might not be updated

Open to further feedback, discussion and correction.

A company needs to monitor the read and write IOPs metrics for their AWS MySQL RDS instance and send real-time alerts to their operations team. Which AWS services can accomplish this? Choose 2 answers
1. Amazon Simple Email Service (Cannot be integrated with CloudWatch directly)
2. Amazon CloudWatch
3. Amazon Simple Queue Service
4. Amazon Route 53
5. Amazon Simple Notification Service
A customer needs to capture all client connection information from their load balancer every five minutes. The company wants to use this data for analyzing traffic patterns and troubleshooting their applications. Which of the following options meets the customer requirements?
1. Enable AWS CloudTrail for the load balancer.
2. Enable access logs on the load balancer. (Refer link)
3. Install the Amazon CloudWatch Logs agent on the load balancer.
4. Enable Amazon CloudWatch metrics on the load balancer (does not provide Client connection information)
A user is running a batch process on EBS backed EC2 instances. The batch process starts a few instances to process Hadoop Map reduce jobs, which can run between 50 – 600 minutes or sometimes for more time. The user wants to configure that the instance gets terminated only when the process is completed. How can the user configure this with CloudWatch?
1. Setup the CloudWatch action to terminate the instance when the CPU utilization is less than 5%
2. Setup the CloudWatch with Auto Scaling to terminate all the instances
3. Setup a job which terminates all instances after 600 minutes
4. It is not possible to terminate instances automatically
A user has two EC2 instances running in two separate regions. The user is running an internal memory management tool, which captures the data and sends it to CloudWatch in US East, using a CLI with the same namespace and metric. Which of the below mentioned options is true with respect to the above statement?
1. The setup will not work as CloudWatch cannot receive data across regions
2. CloudWatch will receive and aggregate the data based on the namespace and metric
3. CloudWatch will give an error since the data will conflict due to two sources
4. CloudWatch will take the data of the server, which sends the data first
A user is sending the data to CloudWatch using the CloudWatch API. The user is sending data 90 minutes in the future. What will CloudWatch do in this case?
1. CloudWatch will accept the data
2. It is not possible to send data of the future
3. It is not possible to send the data manually to CloudWatch
4. The user cannot send data for more than 60 minutes in the future
A user is having data generated randomly based on a certain event. The user wants to upload that data to CloudWatch. It may happen that event may not have data generated for some period due to randomness. Which of the below mentioned options is a recommended option for this case?
1. For the period when there is no data, the user should not send the data at all
2. For the period when there is no data the user should send a blank value
3. For the period when there is no data the user should send the value as 0 (Refer User Guide)
4. The user must upload the data to CloudWatch as having no data for some period will cause an error at CloudWatch monitoring
A user has a weighing plant. The user measures the weight of some goods every 5 minutes and sends data to AWS CloudWatch for monitoring and tracking. Which of the below mentioned parameters is mandatory for the user to include in the request list?
1. Value
2. Namespace (refer put-metric request)
3. Metric Name
4. Timezone
A user has a refrigerator plant. The user is measuring the temperature of the plant every 15 minutes. If the user wants to send the data to CloudWatch to view the data visually, which of the below mentioned statements is true with respect to the information given above?
1. The user needs to use AWS CLI or API to upload the data
2. The user can use the AWS Import Export facility to import data to CloudWatch
3. The user will upload data from the AWS console
4. The user cannot upload data to CloudWatch since it is not an AWS service metric
A user has launched an EC2 instance. The user is planning to setup the CloudWatch alarm. Which of the below mentioned actions is not supported by the CloudWatch alarm?
1. Notify the Auto Scaling launch config to scale up
2. Send an SMS using SNS
3. Notify the Auto Scaling group to scale down
4. Stop the EC2 instance
A user has a refrigerator plant. The user is measuring the temperature of the plant every 15 minutes. If the user wants to send the data to CloudWatch to view the data visually, which of the below mentioned statements is true with respect to the information given above?
1. The user needs to use AWS CLI or API to upload the data
2. The user can use the AWS Import Export facility to import data to CloudWatch
3. The user will upload data from the AWS console
4. The user cannot upload data to CloudWatch since it is not an AWS service metric
A user is trying to aggregate all the CloudWatch metric data of the last 1 week. Which of the below mentioned statistics is not available for the user as a part of data aggregation?
1. Aggregate
2. Sum
3. Sample data
4. Average
A user has setup a CloudWatch alarm on an EC2 action when the CPU utilization is above 75%. The alarm sends a notification to SNS on the alarm state. If the user wants to simulate the alarm action how can he achieve this?
1. Run activities on the CPU such that its utilization reaches above 75%
2. From the AWS console change the state to ‘Alarm’
3. The user can set the alarm state to ‘Alarm’ using CLI
4. Run the SNS action manually
A user is publishing custom metrics to CloudWatch. Which of the below mentioned statements will help the user understand the functionality better?
1. The user can use the CloudWatch Import tool
2. The user should be able to see the data in the console after around 15 minutes
3. If the user is uploading the custom data, the user must supply the namespace, timezone, and metric name as part of the command
4. The user can view as well as upload data using the console, CLI and APIs
An application that you are managing has EC2 instances and DynamoDB tables deployed to several AWS Regions. In order to monitor the performance of the application globally, you would like to see two graphs 1) Avg CPU Utilization across all EC2 instances and 2) Number of Throttled Requests for all DynamoDB tables. How can you accomplish this? [PROFESSIONAL]
1. Tag your resources with the application name, and select the tag name as the dimension in the CloudWatch Management console to view the respective graphs (CloudWatch metrics are regional)
2. Use the CloudWatch CLI tools to pull the respective metrics from each regional endpoint. Aggregate the data offline & store it for graphing in CloudWatch. (Note: CloudWatch cross-account cross-region dashboards and log centralization are now available but this question tests the concept that metrics are regional)
3. Add SNMP traps to each instance and DynamoDB table. Leverage a central monitoring server to capture data from each instance and table. Put the aggregate data into CloudWatch for graphing (Can’t add SNMP traps to DynamoDB as it is a managed service)
4. Add a CloudWatch agent to each instance and attach one to each DynamoDB table. When configuring the agent set the appropriate application name & view the graphs in CloudWatch. (Can’t add agents to DynamoDB as it is a managed service)
You have set up Individual AWS accounts for each project. You have been asked to make sure your AWS Infrastructure costs do not exceed the budget set per project for each month. Which of the following approaches can help ensure that you do not exceed the budget each month? [PROFESSIONAL]
1. Consolidate your accounts so you have a single bill for all accounts and projects (Consolidation will not help limit per account)
2. Set up auto scaling with CloudWatch alarms using SNS to notify you when you are running too many Instances in a given account (many instances do not directly map to cost and would not give exact cost)
3. Set up CloudWatch billing alerts for all AWS resources used by each project, with a notification occurring when the amount for each resource tagged to a particular project matches the budget allocated to the project. (as each project already has a account, no need for resource tagging)
4. Set up CloudWatch billing alerts for all AWS resources used by each account, with email notifications when it hits 50%. 80% and 90% of its budgeted monthly spend
You meet once per month with your operations team to review the past month’s data. During the meeting, you realize that 3 weeks ago, your monitoring system which pings over HTTP from outside AWS recorded a large spike in latency on your 3-tier web service API. You use DynamoDB for the database layer, ELB, EBS, and EC2 for the business logic tier, and SQS, ELB, and EC2 for the presentation layer. Which of the following techniques will NOT help you figure out what happened?
1. Check your CloudTrail log history around the spike’s time for any API calls that caused slowness.
2. Review CloudWatch Metrics graphs to determine which component(s) slowed the system down. (Note: CloudWatch metric data is retained for 15 months. Data at 1-min resolution is available for 15 days, 5-min for 63 days, and 1-hour for 455 days. The 3-week old data would still be available at 5-min or 1-hour resolution.)
3. Review your ELB access logs in S3 to see if any ELBs in your system saw the latency.
4. Analyze your logs to detect bursts in traffic at that time.
You have a high security requirement for your AWS accounts. What is the most rapid and sophisticated setup you can use to react to AWS API calls to your account?
1. Subscription to AWS Config via an SNS Topic. Use a Lambda Function to perform in-flight analysis and reactivity to changes as they occur.
2. Global AWS CloudTrail setup delivering to S3 with an SNS subscription to the deliver notifications, pushing into a Lambda, which inserts records into an ELK stack for analysis.
3. Use a CloudWatch Rule ScheduleExpression to periodically analyze IAM credential logs. Push the deltas for events into an ELK stack and perform ad-hoc analysis there.
4. Amazon EventBridge Rules (formerly CloudWatch Events Rules), which trigger based on all AWS API calls, submitting all events to an AWS Kinesis Stream for arbitrary downstream analysis. (EventBridge (evolution of CloudWatch Events) allows subscription to AWS API calls, and direction of these events into Kinesis Streams. This allows a unified, near real-time stream for all API calls. Refer link)
To monitor API calls against our AWS account by different users and entities, we can use ____ to create a history of calls in bulk for later review, and use ____ for reacting to AWS API calls in real-time.
1. AWS Config; AWS Inspector
2. AWS CloudTrail; AWS Config
3. AWS CloudTrail; Amazon EventBridge (formerly CloudWatch Events) (CloudTrail is a batch API call collection service, EventBridge enables real-time monitoring of calls through the Rules object interface. Refer link)
4. AWS Config; AWS Lambda
You are hired as the new head of operations for a SaaS company. Your CTO has asked you to make debugging any part of your entire operation simpler and as fast as possible. She complains that she has no idea what is going on in the complex, service-oriented architecture, because the developers just log to disk, and it’s very hard to find errors in logs on so many services. How can you best meet this requirement and satisfy your CTO? [PROFESSIONAL]
1. Copy all log files into AWS S3 using a cron job on each instance. Use an S3 Notification Configuration on the <code>PutBucket</code> event and publish events to AWS Lambda. Use the Lambda to analyze logs as soon as they come in and flag issues. (is not fast in search and introduces delay)
2. Begin using CloudWatch Logs on every service. Stream all Log Groups into S3 objects. Use AWS EMR cluster jobs to perform adhoc MapReduce analysis and write new queries when needed. (is not fast in search and introduces delay)
3. Copy all log files into AWS S3 using a cron job on each instance. Use an S3 Notification Configuration on the <code>PutBucket</code> event and publish events to AWS Kinesis. Use Apache Spark on AWS EMR to perform at-scale stream processing queries on the log chunks and flag issues. (is not fast in search and introduces delay)
4. Begin using CloudWatch Logs on every service. Stream all Log Groups into an Amazon OpenSearch Service Domain running Kibana and perform log analysis on a search cluster. (OpenSearch (formerly Elasticsearch) with Kibana/OpenSearch Dashboards is designed for real-time, ad-hoc log analysis and aggregation. Note: CloudWatch now also offers integrated analytics with OpenSearch Service directly.)
Your EC2-Based Multi-tier application includes a monitoring instance that periodically makes application-level read only requests of various application components and if any of those fail more than three times 30 seconds calls CloudWatch to fire an alarm, and the alarm notifies your operations team by email and SMS of a possible application health problem. However, you also need to watch the watcher -the monitoring instance itself – and be notified if it becomes unhealthy. Which of the following is a simple way to achieve that goal? [PROFESSIONAL]
1. Run another monitoring instance that pings the monitoring instance and fires a CloudWatch alarm that notifies your operations team should the primary monitoring instance become unhealthy.
2. Set a CloudWatch alarm based on EC2 system and instance status checks and have the alarm notify your operations team of any detected problem with the monitoring instance.
3. Set a CloudWatch alarm based on the CPU utilization of the monitoring instance and have the alarm notify your operations team if the CPU usage exceeds 50% for more than one minute: then have your monitoring application go into a CPU-bound loop should it detect any application problems.
4. Have the monitoring instances post messages to an SQS queue and then dequeue those messages on another instance should the queue cease to have new messages, the second instance should first terminate the original monitoring instance start another backup monitoring instance and assume the role of the previous monitoring instance and beginning adding messages to the SQS queue.
A company wants to reduce alarm noise in their monitoring system. They have multiple alarms for different components of a web application. Which CloudWatch feature should they use to consolidate multiple alarms into a single notification?
1. CloudWatch Metrics Insights
2. CloudWatch Composite Alarms (Composite alarms combine multiple alarm states using Boolean logic to reduce noise and send notifications only at an aggregated level)
3. CloudWatch Anomaly Detection
4. CloudWatch Dashboards
A DevOps team wants to automatically identify root causes when CloudWatch alarms fire, correlating metrics, logs, and deployment events without manual investigation. Which CloudWatch feature should they use?
1. CloudWatch Logs Insights
2. CloudWatch Contributor Insights
3. CloudWatch Investigations (Powered by Amazon Q Developer, Investigations uses AI to analyze multiple telemetry sources and suggest root causes and remediation steps)
4. CloudWatch Application Signals
An organization wants to send application metrics to CloudWatch using open standards without vendor-specific conversion logic. They need support for high-cardinality metrics with many labels. Which approach should they use?
1. Use CloudWatch PutMetricData API with custom metrics
2. Use CloudWatch Metric Streams to export to a third-party tool
3. Send metrics directly using the OpenTelemetry Protocol (OTLP) to CloudWatch’s native OTel endpoint (CloudWatch supports native OTLP metrics ingestion with up to 150 labels per metric, queryable via PromQL)
4. Use the CloudWatch Agent with StatsD protocol
A company wants to ingest, transform, and route log data from multiple sources to CloudWatch without managing any infrastructure. They need to parse complex log formats and enrich the data during ingestion. Which service should they use?
1. CloudWatch Logs Subscription Filters
2. Amazon Kinesis Data Firehose
3. CloudWatch Pipelines (CloudWatch Pipelines is a fully managed service that ingests, transforms, and routes log data with support for parsing, conditional processing, and AI-assisted processor configuration)
4. CloudWatch Log Transformers only
A team needs to monitor their application’s availability, latency, and error rate with pre-built dashboards and set up Service Level Objectives (SLOs) without building custom monitoring. Which CloudWatch capability provides this?
1. CloudWatch Synthetics
2. CloudWatch Container Insights
3. CloudWatch Internet Monitor
4. CloudWatch Application Signals (Application Signals provides automatic instrumentation, pre-built APM dashboards with golden metrics, and SLO tracking without custom setup)

Amazon CloudWatch Logs – Collection & Insights

May 11, 2018 ~ Last updated on : June 26, 2026 ~ jayendrapatil

Amazon CloudWatch Logs

CloudWatch Logs can be used to monitor, store, and access log files from EC2 instances, CloudTrail, Route 53, and other sources.

CloudWatch Logs uses the log data for monitoring with no code changes required.
CloudWatch Logs requires the unified CloudWatch agent to be installed on EC2 instances and on-premises servers. (Note: The older CloudWatch Logs agent is deprecated and no longer supported. Use the unified CloudWatch agent instead.)

The unified CloudWatch agent can collect both logs and metrics from EC2 instances and on-premises servers.
A VPC endpoint can be configured to keep traffic between VPC and CloudWatch Logs from leaving the Amazon network. It doesn’t require an IGW, NAT, VPN connection, or Direct Connect connection.
CloudWatch Logs allows exporting log data from log groups to an S3 bucket, which can then be used for custom processing and analysis, or to load onto other systems.

Log data is encrypted while in transit and while it is at rest.
Log data can be encrypted using AWS KMS customer managed keys (CMK).
CloudWatch Logs now supports log events up to 1 MB in size (increased from the previous 256 KB limit, as of April 2025).

Required Mainly for SysOps Associate & DevOps Professional Exam

CloudWatch Logs Concepts

Log Events

A log event is a record of some activity recorded by the application or resource being monitored.
Log event record contains two properties: the timestamp of when the event occurred, and the raw event message.

Log Streams

A log stream is a sequence of log events that share the same source for e.g. log events from an Apache access log on a specific host.

Log Groups

Log groups define groups of log streams that share the same retention, monitoring, and access control settings for e.g. Apache access logs from each host grouped through log streams into a single log group.
Each log stream has to belong to one log group.
There is no limit on the number of log streams that can belong to one log group.

Log Classes

CloudWatch Logs offers two log classes:
- Standard – Full-featured log class with real-time monitoring, anomaly detection, Live Tail, Logs Insights, metric filters, subscription filters, and all CloudWatch Logs capabilities.
- Infrequent Access (IA) – Lower-cost log class (50% lower per GB ingestion price) designed for logs that are queried occasionally, such as forensic investigations and compliance auditing. Supports Logs Insights queries, export to S3, encryption, and data protection.

Log class is set at log group creation time and cannot be changed.
Infrequent Access does not support metric filters, subscription filters, Live Tail, or anomaly detection.

Metric Filters

Metric filters can be used to extract metric observations from ingested events and transform them to data points in a CloudWatch metric.

Metric filters are assigned to log groups, and all of the filters assigned to a log group are applied to their log streams.
Metric filters are only available for log groups in the Standard log class.

Retention Settings

Retention settings can be used to specify how long log events are kept in CloudWatch Logs.

Expired log events get deleted automatically.
Retention settings are assigned to log groups, and the retention assigned to a log group is applied to their log streams.

CloudWatch Logs Features

CloudWatch Logs Insights

CloudWatch Logs Insights is an interactive, purpose-built query language for searching and analyzing log data.
Supports queries across multiple log groups and multiple AWS accounts.

Automatically discovers fields in JSON-format logs and generates visualizations.
Supports OpenSearch SQL and PPL (Piped Processing Language) directly within the Logs Insights console for SQL-based log analytics and JOIN operations across log groups (launched Dec 2024).
Field Indexes (Nov 2024) – Index critical log attributes (e.g., requestId, transactionId) to accelerate query performance and reduce scan volume. Up to 20 fields can be indexed per log group. Available at no additional cost.
AI-Powered Natural Language Query Generation (June 2024) – Generate Logs Insights queries using plain language, powered by generative AI.

Scheduled Queries (Nov 2025) – Automate log analysis by running Logs Insights queries on a recurring schedule, delivering results to S3 or EventBridge.
23 New Query Commands and Functions (June 2026) – Includes hash functions (md5, sha256), string functions (strcontains with case-insensitive search, split), conditional logic (if statement), conversion functions (toNumber, toInt, toLong, toDouble), CSV/XML parsing, histogram, and multi-stage pipes for stats.
Facets (June 2026) – Explore log data visually without writing a query for interactive log exploration.

Lookup Tables (June 2026) – Enrich query results with external metadata.
Parameterized Queries (June 2026) – Save reusable query templates with fill-in variables.
JOIN and Sub-queries (June 2026) – Correlate data across log groups in a single query.

Live Tail

Live Tail (launched June 2023) provides real-time streaming of log events as they are ingested into log groups.
Helps quickly view, filter, and troubleshoot issues in incoming logs without writing queries.
Supports fine-grained controls to filter and highlight attributes of interest, and pause/replay logs.

A Live Tail session can last for up to 3 hours.
Available via console, CLI (streaming support added June 2024), and API.
Integrated natively into the AWS Lambda console for real-time function log visibility (Oct 2024).

Only available for Standard log class.

Anomaly Detection and Pattern Analytics

Anomaly Detection (launched Nov 2023) uses machine learning algorithms trained on decades of Amazon and AWS operational data to automatically identify anomalies in log data.
Anomaly detectors train on the past two weeks of log events (training takes up to 15 minutes) and then continuously analyze incoming logs.
Default quota: up to 500 anomaly detectors per account (increased from 10 in Oct 2024).

Pattern Analytics automatically recognizes and clusters patterns among log records, extracting noteworthy content and trends.
Pattern and diff commands use named tokens for easier-to-read results.
Only available for Standard log class.

Log Transformation and Enrichment

Log Transformation (launched Nov 2024) normalizes logs into a consistent format at the time of ingestion.

Supports pre-configured templates for common AWS services (WAF, VPC Flow Logs) and custom transformers (Grok patterns).
Each transformer can extract up to 200 fields from a log event.
Transformed logs can be leveraged for field indexes, discovered fields in Logs Insights, metric filters, and subscription filters.

Included with existing Standard log class ingestion price (no additional cost for transformation).
Supports OCSF (Open Cybersecurity Schema Framework) and OTel format conversion for security and compliance data.

Log Analytics (Unified Console)

Log Analytics (launched June 2026) is a unified console experience that brings together:
- CloudWatch Logs Insights for querying and analyzing log data
- Live Tail for real-time log streaming
- Contributor Insights for identifying top contributors

Supports multiple queries in different tabs simultaneously.
Combines facets, saved queries with parameters, natural language query generation, and visualizations in a single interface.

CloudWatch Logs Use Cases

Monitor Logs from EC2 Instances in Real-time

Can help monitor applications and systems using log data.
Can help track number of errors for e.g. 404, 500, or even specific literal terms “NullReferenceException”, occurring in applications, which can then be matched to a threshold to send notification.

Monitor AWS CloudTrail Logged Events

Can be used to monitor particular API activity as captured by CloudTrail by creating alarms in CloudWatch and receiving notifications.

Archive Log Data

Can help store log data in highly durable storage, an alternative to S3.
Log retention setting can be modified, so that any log events older than this setting are automatically deleted.

Log Route 53 DNS Queries

Can help log information about the DNS queries that Route 53 receives.

Managed Syslog Ingestion

Managed Syslog Ingestion (launched June 2026) enables sending syslog messages from firewalls, routers, switches, and Linux servers directly into CloudWatch Logs.
Eliminates the need for custom syslog collection infrastructure.

Real-time Processing of Log Data with Subscriptions

Subscriptions provide access to a real-time feed of log events from CloudWatch Logs and deliver them to other services such as Kinesis Data Streams, Amazon Data Firehose, AWS Lambda, or Amazon OpenSearch Service for custom processing, analysis, or loading to other systems.
A subscription filter defines the filter pattern to use for filtering which log events get delivered, as well as the destination for matching log events.

CloudWatch Logs log group can be configured to stream data to an Amazon OpenSearch Service cluster in near real-time. (Note: Amazon Elasticsearch Service was renamed to Amazon OpenSearch Service in September 2021.)
Account-Level Subscription Filters (Jan 2024) – Deliver real-time log events from all log groups in an account to a Kinesis Data Stream, Amazon Data Firehose, or Lambda using a single account-level filter (one per account per Region).

Cross-Account and Cross-Region Log Centralization

Cross-Account Observability enables searching log groups across multiple accounts, running cross-account Logs Insights queries, and creating Contributor Insights rules.

Cross-Account and Cross-Region Log Centralization (Sept 2025) copies log data from multiple AWS accounts and regions into a single destination account for centralized management.

Searching and Filtering

CloudWatch Logs allows searching and filtering the log data by creating one or more metric filters.
Metric filters define the terms and patterns to look for in log data as it is sent to CloudWatch Logs.

CloudWatch Logs uses these metric filters to turn log data into numerical CloudWatch metrics that can be graphed or used to set alarms.

AWS Certification Exam Practice Questions

Questions are collected from Internet and the answers are marked as per my knowledge and understanding (which might differ with yours).

AWS services are updated everyday and both the answers and questions might be outdated soon, so research accordingly.

AWS exam questions are not updated to keep up the pace with AWS updates, so even if the underlying feature has changed the question might not be updated

Open to further feedback, discussion and correction.

Once we have our logs in CloudWatch, we can do a number of things such as: Choose 3. Choose the 3 correct answers:[CDOP]
1. Send the log data to AWS Lambda for custom processing or to load into other systems
2. Stream the log data to Amazon Kinesis
3. Stream the log data into Amazon OpenSearch Service in near real-time with CloudWatch Logs subscriptions.
4. Record API calls for your AWS account and delivers log files containing API calls to your Amazon S3 bucket
You have decided to set the threshold for errors on your application to a certain number and once that threshold is reached you need to alert the Senior DevOps engineer. What is the best way to do this? Choose 3. Choose the 3 correct answers: [CDOP]
1. Set the threshold your application can tolerate in a CloudWatch Logs group and link a CloudWatch alarm on that threshold.
2. Use the unified CloudWatch agent to send log data from the app to CloudWatch Logs from Amazon EC2 instances
3. Pipe data from EC2 to the application logs using AWS Data Pipeline and CloudWatch
4. Once a CloudWatch alarm is triggered, use SNS to notify the Senior DevOps Engineer.
You are hired as the new head of operations for a SaaS company. Your CTO has asked you to make debugging any part of your entire operation simpler and as fast as possible. She complains that she has no idea what is going on in the complex, service-oriented architecture, because the developers just log to disk, and it’s very hard to find errors in logs on so many services. How can you best meet this requirement and satisfy your CTO? [CDOP]
1. Copy all log files into AWS S3 using a cron job on each instance. Use an S3 Notification Configuration on the PutBucket event and publish events to AWS Lambda. Use the Lambda to analyze logs as soon as they come in and flag issues. (is not fast in search and introduces delay)
2. Begin using CloudWatch Logs on every service. Stream all Log Groups into S3 objects. Use AWS EMR cluster jobs to perform adhoc MapReduce analysis and write new queries when needed. (is not fast in search and introduces delay)
3. Copy all log files into AWS S3 using a cron job on each instance. Use an S3 Notification Configuration on the PutBucket event and publish events to AWS Kinesis. Use Apache Spark on AWS EMR to perform at-scale stream processing queries on the log chunks and flag issues. (is not fast in search and introduces delay)
4. Begin using CloudWatch Logs on every service. Stream all Log Groups into an Amazon OpenSearch Service domain running OpenSearch Dashboards and perform log analysis on a search cluster. (OpenSearch with Dashboards (formerly ELK – Elasticsearch, Kibana stack) is designed specifically for real-time, ad-hoc log analysis and aggregation)

You use Amazon CloudWatch as your primary monitoring system for your web application. After a recent software deployment, your users are getting Intermittent 500 Internal Server Errors when using the web application. You want to create a CloudWatch alarm, and notify an on-call engineer when these occur. How can you accomplish this using AWS services? (Choose three.) [CDOP]
1. Deploy your web application as an AWS Elastic Beanstalk application. Use the default Elastic Beanstalk CloudWatch metrics to capture 500 Internal Server Errors. Set a CloudWatch alarm on that metric.
2. Install the unified CloudWatch agent on your servers to stream web application logs to CloudWatch.
3. Use Amazon Simple Email Service to notify an on-call engineer when a CloudWatch alarm is triggered.
4. Create a CloudWatch Logs group and define metric filters that capture 500 Internal Server Errors. Set a CloudWatch alarm on that metric.
5. Use Amazon Simple Notification Service to notify an on-call engineer when a CloudWatch alarm is triggered.
6. Use AWS Data Pipeline to stream web application logs from your servers to CloudWatch.
A company needs to consolidate logs from multiple AWS accounts for compliance auditing but wants to minimize costs since logs are only queried during investigations. Which CloudWatch Logs approach is most cost-effective? [SAA/SysOps]
1. Use Standard log class with metric filters for all log groups
2. Use Infrequent Access log class for the log groups and query with Logs Insights when needed
3. Export all logs to S3 immediately and use Athena for queries
4. Use Live Tail to monitor logs in real-time across accounts

A DevOps team wants to quickly identify anomalous log patterns after a deployment without writing custom detection rules. Which CloudWatch Logs feature should they use? [CDOP]
1. Metric filters with CloudWatch Alarms
2. CloudWatch Logs Insights scheduled queries
3. CloudWatch Logs Anomaly Detection with pattern analytics
4. Subscription filters to Lambda for custom processing
A team needs to stream log events from ALL log groups in their AWS account to a central Kinesis Data Stream without creating individual subscription filters for each log group. What should they use? [CDOP]
1. Cross-account observability with a monitoring account
2. CloudWatch Logs transformation with Kinesis destination
3. Account-level subscription filter policy
4. CloudWatch Logs Live Tail with CLI streaming
An engineer wants to speed up CloudWatch Logs Insights queries that frequently filter by requestId across large log groups. Which feature should they configure? [CDOP/SysOps]
1. Log transformation with Grok patterns
2. Field indexes on the requestId field
3. Infrequent Access log class for the log group
4. Scheduled queries with S3 delivery

References

CloudWatch Monitoring Supported AWS Services

May 31, 2016 ~ Last updated on : June 21, 2026 ~ jayendrapatil ~ 8 Comments

CloudWatch Monitoring Supported AWS Services

CloudWatch offers either basic or detailed monitoring for supported AWS services.

Basic monitoring means that a service sends data points to CloudWatch every five minutes.
Detailed monitoring means that a service sends data points to CloudWatch every minute.

If the AWS service supports both basic and detailed monitoring, the basic would be enabled by default and the detailed monitoring needs to be enabled for detailed metrics.
High-Resolution Custom Metrics allow publishing data at 1-second resolution using the PutMetricData API with a StorageResolution of 1.

Monitoring Categories

Basic Monitoring – Free, default set of metrics published at 5-minute intervals for most services.

Detailed Monitoring – Paid, more frequent metrics (typically 1-minute intervals). Must be explicitly enabled.
High-Resolution Custom Metrics – Custom metrics published at up to 1-second intervals using PutMetricData API or Embedded Metric Format (EMF).

Services Offering Detailed Monitoring

The following services officially offer detailed monitoring (paid, more fine-grained metrics):

Amazon API Gateway – Additional dimensions for detailed metrics
AWS AppSync – Detailed CloudWatch metrics
Amazon CloudFront – Additional distribution metrics
Amazon EC2 – 1-minute metrics (vs. 5-minute basic)
AWS Elastic Beanstalk – Enhanced health reporting and monitoring
Amazon Kinesis Data Streams – Enhanced shard-level metrics
AWS Lambda – Event source mapping metrics
Amazon Managed Streaming for Apache Kafka (MSK) – Per-broker, per-topic metrics
Amazon S3 – Request metrics at 1-minute intervals
Amazon SES – Detailed monitoring via event publishing

AWS Services with Monitoring Support

Auto Scaling
- By default, basic monitoring is enabled when the launch configuration is created using the AWS Management Console, and detailed monitoring is enabled when the launch configuration is created using the AWS CLI or an API.
- Auto Scaling sends data to CloudWatch every 5 minutes by default when created from Console.
- For an additional charge, you can enable detailed monitoring for Auto Scaling, which sends data to CloudWatch every minute.
Amazon CloudFront
- Amazon CloudFront sends data to CloudWatch every minute by default.
- Additional distribution metrics (detailed monitoring) can be enabled for more fine-grained visibility.
Amazon CloudSearch
- Amazon CloudSearch sends data to CloudWatch every minute by default.
Amazon EventBridge (formerly Amazon CloudWatch Events)
- Amazon EventBridge sends data to CloudWatch every minute by default.

Amazon CloudWatch Logs
- Amazon CloudWatch Logs sends data to CloudWatch every minute by default.
Amazon DynamoDB
- Amazon DynamoDB sends data to CloudWatch every minute for some metrics and every 5 minutes for other metrics.
- DynamoDB Contributor Insights provides additional metrics for table and global secondary index access patterns.
Amazon Elastic Container Service (Amazon ECS)
- Amazon ECS sends data to CloudWatch every minute.
- Container Insights provides additional detailed metrics at the cluster, service, task, and container level including CPU, memory, network, and storage metrics.
Amazon ElastiCache
- Amazon ElastiCache sends data to CloudWatch every minute.
Amazon Elastic Block Store (EBS)
- Amazon EBS sends data to CloudWatch every 5 minutes for gp2, st1, and sc1 volumes.
- Provisioned IOPS SSD (io1 and io2) volumes automatically send one-minute metrics to CloudWatch.
- gp3 volumes also send metrics at 1-minute intervals.
Amazon Elastic Compute Cloud (EC2)
- Amazon EC2 sends data to CloudWatch every 5 minutes by default. For an additional charge, you can enable detailed monitoring for Amazon EC2, which sends data to CloudWatch every minute.
Elastic Load Balancing
- Elastic Load Balancing sends data to CloudWatch every minute (applies to ALB, NLB, GLB, and Classic Load Balancer).

Amazon EMR (formerly Amazon Elastic MapReduce)
- Amazon EMR sends basic data to CloudWatch every 5 minutes by default at no additional cost.
- Starting with Amazon EMR Release 7.0+, the CloudWatch Agent can publish 34 enhanced metrics every minute (additional charges apply).
- EMR Serverless sends metrics to CloudWatch every minute.

Amazon OpenSearch Service (formerly Amazon Elasticsearch Service)
- Amazon OpenSearch Service sends data to CloudWatch every minute.
Amazon Kinesis Data Streams (formerly Amazon Kinesis Streams)
- Amazon Kinesis Data Streams sends stream-level data to CloudWatch every minute.
- Enhanced shard-level metrics (detailed monitoring) provide additional per-shard metrics.

Amazon Data Firehose (formerly Amazon Kinesis Data Firehose)
- Amazon Data Firehose sends data to CloudWatch every minute.
AWS Lambda
- AWS Lambda sends data to CloudWatch every minute.
- Lambda Insights provides enhanced monitoring with system-level metrics (CPU, memory, network) at 1-minute intervals.

Amazon SageMaker AI
- Amazon SageMaker AI (which replaced the legacy Amazon Machine Learning service) sends training, endpoint, and transform job metrics to CloudWatch every minute.

⚠️ Note: The original Amazon Machine Learning service is no longer accepting new users. AWS recommends using Amazon SageMaker AI for machine learning workloads.

Amazon Redshift
- Amazon Redshift sends data to CloudWatch every minute.

Amazon Relational Database Service (RDS)
- Amazon RDS sends data to CloudWatch every minute.
- CloudWatch Database Insights (launched Dec 2024) provides comprehensive database observability with fleet-level and instance-level dashboards.
Amazon Route 53
- Amazon Route 53 sends data to CloudWatch every minute.

Amazon Simple Notification Service (SNS)
- Amazon SNS sends data to CloudWatch every 5 minutes.
- SNS does not support detailed (1-minute) monitoring.
Amazon Simple Queue Service (SQS)
- Amazon SQS sends data to CloudWatch every 5 minutes.
Amazon Simple Storage Service (S3)
- Amazon S3 sends storage metrics (bucket size, object count) to CloudWatch once a day (basic monitoring, free).
- Request metrics (detailed monitoring) are available at 1-minute intervals and are billed as CloudWatch custom metrics.
- 1-minute metrics are available at the bucket-level by default when request metrics are enabled.

Amazon Simple Workflow Service (SWF)
- Amazon SWF sends data to CloudWatch every 5 minutes.
- Note: AWS Step Functions is the recommended alternative for new workflow orchestration workloads.
AWS Storage Gateway
- AWS Storage Gateway sends data to CloudWatch every 5 minutes.
AWS WAF
- AWS WAF sends data to CloudWatch every minute.
Amazon WorkSpaces
- Amazon WorkSpaces sends data to CloudWatch every 5 minutes.

⚠️ AWS OpsWorks – End of Life

AWS OpsWorks reached End of Life (EOL) on May 26, 2024. The service has been disabled for both new and existing customers. The OpsWorks console, API, CLI, and CloudFormation resources are no longer available.

Alternatives: AWS Systems Manager, AWS CodeDeploy, AWS CloudFormation

Additional Services Publishing CloudWatch Metrics (2024-2026)

The following additional AWS services publish metrics to CloudWatch (not in the original list):

Amazon API Gateway – Sends metrics every minute
AWS AppSync – Sends metrics every minute
Amazon EKS – Control plane metrics and Container Insights
Amazon Bedrock – Model invocation and throughput metrics
AWS Step Functions – Execution metrics every minute
Amazon Aurora – Database metrics every minute (with Database Insights)

AWS Fargate – Container-level metrics via Container Insights
Amazon MSK – Streaming metrics with per-broker/topic detail
AWS Network Firewall – Firewall metrics every minute
Amazon MemoryDB – Database metrics every minute

CloudWatch Enhanced Observability Features

Container Insights – Collects and aggregates metrics and logs from containerized applications on Amazon ECS, Amazon EKS, and Kubernetes. Provides cluster, node, pod, task, and service level metrics.
Lambda Insights – Enhanced monitoring for Lambda functions with system-level metrics (CPU, memory, network, disk).
Database Insights (Dec 2024) – Comprehensive database observability for Amazon RDS and Aurora with fleet-level health monitoring and instance-level SQL query analysis.

Application Signals (June 2024) – Application performance monitoring (APM) with pre-built dashboards showing volume, availability, latency, faults, and errors.
Internet Monitor – Near-continuous internet measurements for availability and performance, tailored to your workload footprint on AWS.
CloudWatch Investigations – AI-powered investigation of operational issues across services.

AWS Certification Exam Practice Questions

Questions are collected from Internet and the answers are marked as per my knowledge and understanding (which might differ with yours).

AWS services are updated everyday and both the answers and questions might be outdated soon, so research accordingly.

AWS exam questions are not updated to keep up the pace with AWS updates, so even if the underlying feature has changed the question might not be updated

Open to further feedback, discussion and correction.

What is the minimum time interval for the data that Amazon CloudWatch receives and aggregates?
1. One second (High-resolution custom metrics support 1-second resolution)
2. Five seconds
3. One minute
4. Three minutes
5. Five minutes
Note: The original answer was “One minute” which was correct for standard metrics. With high-resolution custom metrics (introduced 2017), CloudWatch supports 1-second resolution. Exam questions may still reference 1 minute as the minimum for AWS service metrics.
In the ‘Detailed’ monitoring data available for your Amazon EBS volumes, Provisioned IOPS volumes automatically send _____ minute metrics to Amazon CloudWatch.
1. 3
2. 1
3. 5
4. 2

Using Amazon CloudWatch’s Free Tier, what is the frequency of metric updates, which you receive?
1. 5 minutes
2. 500 milliseconds.
3. 30 seconds
4. 1 minute
What is the type of monitoring data (for Amazon EBS volumes) which is available automatically in 5-minute periods at no charge called?
1. Basic
2. Primary
3. Detailed
4. Local
A user has created an Auto Scaling group using CLI. The user wants to enable CloudWatch detailed monitoring for that group. How can the user configure this?
1. When the user sets an alarm on the Auto Scaling group, it automatically enables detail monitoring
2. By default detailed monitoring is enabled for Auto Scaling (Detailed monitoring is enabled when you create the launch configuration using the AWS CLI or an API)
3. Auto Scaling does not support detailed monitoring
4. Enable detail monitoring from the AWS console
A user is trying to understand the detailed CloudWatch monitoring concept. Which of the below mentioned services provides detailed monitoring with CloudWatch without charging the user extra?
1. AWS Auto Scaling
2. AWS Route 53
3. AWS EMR
4. AWS SNS

A user is trying to understand the detailed CloudWatch monitoring concept. Which of the below mentioned services does not provide detailed monitoring with CloudWatch?
1. AWS EMR (EMR sends basic metrics every 5 minutes by default; enhanced monitoring at 1-minute intervals is available starting with EMR 7.0+ via CloudWatch Agent)
2. AWS RDS
3. AWS ELB
4. AWS Route53
A user has enabled detailed CloudWatch monitoring with the AWS Simple Notification Service. Which of the below mentioned statements helps the user understand detailed monitoring better?
1. SNS will send data every minute after configuration
2. There is no need to enable since SNS provides data every minute
3. AWS CloudWatch does not support monitoring for SNS
4. SNS cannot provide data every minute
A user has configured an Auto Scaling group with ELB. The user has enabled detailed CloudWatch monitoring on Auto Scaling. Which of the below mentioned statements will help the user understand the functionality better?
1. It is not possible to setup detailed monitoring for Auto Scaling
2. In this case, Auto Scaling will send data every minute and will charge the user extra
3. Detailed monitoring will send data every minute without additional charges
4. Auto Scaling sends data every minute only and does not charge the user

Which of the following CloudWatch monitoring features provides near real-time visibility into application performance with pre-built dashboards?
1. CloudWatch Logs Insights
2. CloudWatch Alarms
3. CloudWatch Application Signals
4. CloudWatch Contributor Insights
What is the minimum resolution supported by CloudWatch high-resolution custom metrics?
1. 5 seconds
2. 10 seconds
3. 30 seconds
4. 1 second
Which CloudWatch feature provides comprehensive database observability with fleet-level health monitoring for Amazon RDS and Aurora?
1. CloudWatch Logs Insights
2. Enhanced Monitoring
3. Performance Insights
4. CloudWatch Database Insights

References

AWS ELB Monitoring

March 18, 2016 ~ Last updated on : June 23, 2026 ~ jayendrapatil ~ 9 Comments

AWS ELB Monitoring

Elastic Load Balancing publishes data points to Amazon CloudWatch about the load balancers and targets (or back-end instances for Classic Load Balancer).

Elastic Load Balancing reports metrics to CloudWatch only when requests are flowing through the load balancer.
- If there are requests flowing through the load balancer, Elastic Load Balancing measures and sends its metrics in 60-second intervals.
- If there are no requests flowing through the load balancer or no data for a metric, the metric is not reported.
AWS provides four types of load balancers, each with its own monitoring capabilities:
- Application Load Balancer (ALB) – Layer 7, HTTP/HTTPS/gRPC
- Network Load Balancer (NLB) – Layer 4, TCP/UDP/TLS
- Gateway Load Balancer (GWLB) – Layer 3, transparent network gateway
- Classic Load Balancer (CLB) – Previous generation (Layer 4/7)

ELB monitoring options include CloudWatch metrics, access logs, connection logs, health check logs, CloudTrail logs, and CloudWatch Internet Monitor.

CloudWatch Metrics

Classic Load Balancer (CLB) Metrics

CLB metrics use the AWS/ELB namespace.
HealthyHostCount, UnHealthyHostCount
- Number of healthy and unhealthy instances registered with the load balancer.
- Most useful statistics are Average, Min, and Max.

RequestCount
- Number of requests completed or connections made during the specified interval (1 or 5 minutes).
- Most useful statistic is Sum.
Latency
- Time elapsed, in seconds, after the request leaves the load balancer until the headers of the response are received.
- Most useful statistic is Average.
SurgeQueueLength
- Total number of requests that are pending routing.
- Load balancer queues a request if it is unable to establish a connection with a healthy instance in order to route the request.
- Maximum size of the queue is 1,024. Additional requests are rejected when the queue is full.
- Most useful statistic is Max, because it represents the peak of queued requests.
SpilloverCount
- The total number of requests that were rejected because the surge queue is full. Should ideally be 0.
- Most useful statistic is Sum.
HTTPCode_ELB_4XX, HTTPCode_ELB_5XX
- Client and server error codes generated by the load balancer.
- Most useful statistic is Sum.

HTTPCode_Backend_2XX, HTTPCode_Backend_3XX, HTTPCode_Backend_4XX, HTTPCode_Backend_5XX
- Number of HTTP response codes generated by registered instances.
- Most useful statistic is Sum.

Application Load Balancer (ALB) Metrics

ALB metrics use the AWS/ApplicationELB namespace.
ActiveConnectionCount – Total concurrent TCP connections active from clients to the load balancer and from the load balancer to targets. Useful statistic: Sum.
NewConnectionCount – Total new TCP connections established from clients to the load balancer and from the load balancer to targets. Useful statistic: Sum.

RejectedConnectionCount – Number of connections rejected because the load balancer reached its maximum number of connections. Useful statistic: Sum.
RequestCount – Number of requests processed over IPv4 and IPv6. Useful statistic: Sum.
TargetResponseTime – Time elapsed after the request leaves the load balancer until the target starts to send response headers. Useful statistics: Average, pNN.NN (percentiles).

HealthyHostCount, UnHealthyHostCount – Number of healthy/unhealthy targets. Useful statistics: Average, Min, Max.
HTTPCode_Target_2XX_Count through 5XX_Count – HTTP response codes generated by targets. Useful statistic: Sum.
HTTPCode_ELB_4XX_Count, HTTPCode_ELB_5XX_Count – HTTP error codes generated by the load balancer itself. Useful statistic: Sum.

ClientTLSNegotiationErrorCount – TLS connections initiated by clients that did not establish a session with the load balancer. Useful statistic: Sum.
TargetConnectionErrorCount – Connections that were not successfully established between the load balancer and target. Useful statistic: Sum.
ProcessedBytes – Total bytes processed by the load balancer over IPv4 and IPv6. Useful statistic: Sum.

ConsumedLCUs – Number of Load Balancer Capacity Units (LCU) consumed. Used for billing calculations.
RuleEvaluations – Number of rules evaluated while processing requests.
AnomalousHostCount – Number of targets detected with anomalies (used with Automatic Target Weights). Useful statistics: Min, Max.

Network Load Balancer (NLB) Metrics

NLB metrics use the AWS/NetworkELB namespace.
ActiveFlowCount – Total number of concurrent flows (connections) from clients to targets. Useful statistic: Average.

NewFlowCount – Total number of new flows established from clients to targets. Useful statistic: Sum.
ProcessedBytes – Total bytes processed by the load balancer (TCP/TLS, UDP). Useful statistic: Sum.
TCP_Client_Reset_Count, TCP_Target_Reset_Count, TCP_ELB_Reset_Count – Number of reset (RST) packets sent from client, target, or the load balancer.

HealthyHostCount, UnHealthyHostCount – Number of healthy/unhealthy targets.
ConsumedLCUs – Number of Network Load Balancer Capacity Units consumed.
PeakBytesPerSecond – Highest average bytes per second for the load balancer during a period.

Gateway Load Balancer (GWLB) Metrics

GWLB metrics use the AWS/GatewayELB namespace.
ActiveFlowCount, NewFlowCount – Concurrent and new flows from clients to targets.

ProcessedBytes – Total bytes processed by the GWLB.
HealthyHostCount, UnHealthyHostCount – Number of healthy/unhealthy targets.
GWLB does NOT generate access logs since it is a transparent Layer 3 load balancer that does not terminate flows.

Elastic Load Balancer Access Logs

Elastic Load Balancing provides access logs that capture detailed information about all requests sent to the load balancer.
Each log contains information such as the time the request was received, the client’s IP address, latencies, request paths, and server responses.

Access logging is disabled by default and can be enabled without any additional charge. You are only charged for S3 storage.
Access logs are supported for ALB, NLB, and CLB. GWLB does not generate access logs.

ALB Access Logs

ALB publishes a log file for each load balancer node every 5 minutes to Amazon S3.

Log entries include: request type, timestamp, ELB name, client:port, target:port, request processing time, target processing time, response processing time, ELB/target status codes, received/sent bytes, request details, user agent, SSL cipher/protocol, target group ARN, trace ID, and more.

NLB Access Logs

NLB access logs capture information about TLS requests sent to the load balancer.
Logs can be stored in Amazon S3.
New (Nov 2025): NLB access logs now support delivery as CloudWatch Vended Logs, enabling direct delivery to CloudWatch Logs, Amazon Data Firehose, and Amazon S3 with Apache Parquet format support. This allows real-time log analysis using CloudWatch Logs Insights and Live Tail.

ALB Connection Logs

Connection logs capture detailed information about TLS connections established between clients and the ALB.
Useful for troubleshooting TLS client connection issues (e.g., mTLS failures, cipher mismatches).
Connection logs are stored in Amazon S3, with a log file published every 5 minutes.

This is an optional feature, disabled by default.
Log entries include: timestamp, client IP:port, listener port, TLS protocol/cipher, connection status, client certificate details (for mTLS), and more.

ALB Health Check Logs

New (Nov 2025): ALB now supports Health Check Logs that send detailed target health check data directly to a designated Amazon S3 bucket.
This optional feature captures:
- Health check status (healthy/unhealthy)
- Timestamps
- Target identification data
- Failure reasons for unhealthy targets

Health check logs are published every 5 minutes per load balancer node.
Helps troubleshoot intermittent target health check failures without needing to rely solely on CloudWatch metrics.
No additional charge; you pay only for S3 storage.

CloudWatch Internet Monitor

Amazon CloudWatch Internet Monitor provides internet performance and availability measurements for user traffic to load balancers.

Monitors internet traffic patterns and identifies issues that affect internet connectivity between users and AWS.
Supported for both ALB and NLB.
NLB integration (Sep 2024): You can create or associate a monitor for an NLB directly when creating it in the AWS Management Console.
Provides city-level visibility into performance impairments and their geographic scope.

CloudWatch Network Flow Monitor

New (Dec 2024, re:Invent): CloudWatch Network Flow Monitor offers network performance monitoring across AWS managed services.
Provides near real-time visibility into network performance for traffic between compute resources (EC2, EKS), to AWS services (S3, DynamoDB), and to other AWS Regions.
Uses lightweight agents to gather TCP connection performance statistics (packet loss, latency).

Can determine if AWS is the cause of a detected network issue for monitored flows.

ALB Automatic Target Weights (ATW)

New (Nov 2023): ALB supports Automatic Target Weights (ATW), which uses anomaly detection to optimize traffic routing.
ATW detects and mitigates gray failures — situations where a target passes health checks but still returns elevated errors.

Anomaly detection is automatically enabled on HTTP/HTTPS target groups with at least three healthy targets.
ATW analyzes HTTP return status codes and TCP/TLS errors to identify anomalous targets and reduces traffic to them.
Provides the AnomalousHostCount CloudWatch metric to monitor detected anomalies.

CloudWatch Anomaly Detection Alarms

CloudWatch anomaly detection uses machine learning to model expected metric behavior and automatically creates upper and lower bounds.

Can be used with ELB metrics like TargetResponseTime, RequestCount, HTTPCode_ELB_5XX to detect unusual patterns.
Recommended approach for monitoring ELB performance without manually setting static thresholds.
Works with ALB, NLB, CLB, and GWLB metrics.

CloudTrail Logs

AWS CloudTrail captures all API calls to the Elastic Load Balancing API made by or on behalf of your AWS account.

API calls can be made directly, or indirectly through the AWS Management Console, AWS CLI, or SDKs.
CloudTrail stores the information as log files in an Amazon S3 bucket.
Logs can be used to monitor load balancer activity and determine what API call was made, what source IP address was used, who made the call, when it was made, and so on.

Applies to all ELB types (ALB, NLB, GWLB, CLB).

Classic Load Balancer – Migration Recommendation

⚠️ Note: Classic Load Balancer is the previous generation load balancer. AWS strongly recommends migrating to Application Load Balancer (Layer 7) or Network Load Balancer (Layer 4).

EC2-Classic networking was fully retired in August 2023. While CLB continues to function in VPC, no new features are being added to it. Use the AWS Migration Wizard to move to ALB or NLB.

AWS Certification Exam Practice Questions

Questions are collected from Internet and the answers are marked as per my knowledge and understanding (which might differ with yours).

AWS services are updated everyday and both the answers and questions might be outdated soon, so research accordingly.

AWS exam questions are not updated to keep up the pace with AWS updates, so even if the underlying feature has changed the question might not be updated

Open to further feedback, discussion and correction.

An admin is planning to monitor the ELB. Which of the below mentioned services does not help the admin capture the monitoring information about the ELB activity?
1. ELB Access logs
2. ELB health check
3. CloudWatch metrics
4. ELB API calls with CloudTrail

A customer needs to capture all client connection information from their load balancer every five minutes. The company wants to use this data for analyzing traffic patterns and troubleshooting their applications. Which of the following options meets the customer requirements?
1. Enable AWS CloudTrail for the load balancer.
2. Enable access logs on the load balancer.
3. Install the Amazon CloudWatch Logs agent on the load balancer.
4. Enable Amazon CloudWatch metrics on the load balancer.
Your supervisor has requested a way to analyze traffic patterns for your application. You need to capture all connection information from your load balancer every 10 minutes. Pick a solution from below. Choose the correct answer:
1. Enable access logs on the load balancer.
2. Create a custom metric CloudWatch filter on your load balancer.
3. Use a CloudWatch Logs Agent.
4. Use AWS CloudTrail with your load balancer.
A company runs a web application behind an Application Load Balancer. Some users are experiencing intermittent 5XX errors but health checks show all targets as healthy. Which ALB feature can automatically detect and mitigate this issue?
1. Cross-Zone Load Balancing
2. Automatic Target Weights (ATW)
3. Connection Draining
4. Sticky Sessions
A DevOps engineer needs to troubleshoot why targets behind an ALB are intermittently failing health checks. Which recently introduced feature provides detailed health check failure reasons stored in S3?
1. ALB Access Logs
2. CloudWatch HealthyHostCount metric
3. ALB Health Check Logs
4. AWS CloudTrail

A solutions architect wants to analyze NLB access logs in near real-time using CloudWatch Logs Insights. Which delivery option should they configure?
1. Enable NLB access logs to S3 and create Athena queries
2. Configure NLB access logs as CloudWatch Vended Logs
3. Enable VPC Flow Logs on the NLB
4. Install CloudWatch Agent on NLB nodes
Which of the following is a metric specific to Classic Load Balancer that indicates the load balancer cannot route requests because the queue is full?
1. RejectedConnectionCount
2. TargetConnectionErrorCount
3. SpilloverCount
4. HTTPCode_ELB_503

A company wants to identify if AWS infrastructure is causing latency issues for users connecting to their Network Load Balancer from different geographic locations. Which service should they use?
1. AWS X-Ray
2. CloudWatch Metrics
3. Amazon CloudWatch Internet Monitor
4. VPC Flow Logs