AWS
StackState Self-hosted v4.6.x
Last updated
StackState Self-hosted v4.6.x
Last updated
This page describes StackState version 4.6.
Amazon Web Services (AWS) is a major cloud provider. This StackPack enables in-depth monitoring of AWS services.
StackState Agent V2 collects all service responses from the target AWS account.
Topology is updated in real time:
Once an hour, all services are queried to gain a full point-in-time snapshot of resources.
Once a minute, Cloudtrail and Eventbridge events are read to find changes to resources.
Logs are retrieved once a minute from CloudWatch and a central S3 bucket. These are mapped to associated components in StackState.
Metrics are retrieved on-demand by the StackState CloudWatch plugin. These are mapped to associated components in StackState.
VPC FlowLogs are retrieved once a minute from the configured S3 bucket. Private network traffic inside VPCs is analysed to create relations between EC2 and RDS database components in StackState.
Communication between NodeJS Lambda functions and the AWS services that they communicate with is monitored using OpenTelemetry traces.
AWS is a StackState core integration.
To set up the StackState AWS integration, you need to have:
StackState Agent V2 installed on a machine which can connect to both AWS and StackState.
AWS CLI version 2.0.4 or later is installed on the environment where StackState is running.
The following AWS accounts:
At least one target AWS account that will be monitored.
An AWS account for the StackState Agent to use when retrieving data from the target AWS accounts. It is recommended to use a separate shared account for this and not use any of the accounts that will be monitored by StackState.
It is recommended to have two different AWS accounts: One that is being monitored (the monitor account) and another for the StackState Agent (the Agent account).
Monitor account - used to deploy a CloudFormation Stack. The cloudFormation stack will create an IAM role that has the permissions required to retrieve data from this monitor account (StackStateAwsIntegrationRole).
Agent account - used to retrieve data from the monitor account. StackState Agent must have permissions to assume the role StackStateAwsIntegrationRole created by the monitor account's CloudFormation Stack. This can come from any of the following:
An IAM role attached to the EC2 instance where the Agent runs.
An IAM role attached to the EKS pod where the Cluster Agent runs.
An AWS user configured in the AWS check on the Agent.
The IAM role of the Agent account queries AWS data from the monitor account by assuming the role StackStateAwsIntegrationRole. This data is then returned to the StackState Agent where it is processed and sent on to StackState.
StackState pulls CloudWatch metrics directly from AWS. If StackState is running within AWS (EC2 or EKS), an IAM role can be used for authentication when collecting metrics from CloudWatch:
In AWS, Create the required policy and attach it to the relevant IAM role:
When you install an AWS StackPack instance, set the following parameter values:
AWS Access Key ID: use-role
AWS Secret Access Key: use-role
Note that StackState Agent also connects to AWS to pull topology and events data. The Agent can also be configured to authenticate using an IAM role, see IAM role for StackState Agent on EC2 or EKS.
If StackState Agent runs in an AWS environment, an IAM role can be attached to the EC2 instance or EKS pod that it runs on. The Agent will then use this role by default.
If the Agent runs in an EKS pod: In AWS, create the required policy and attach it to the relevant IAM role - Agent IAM role: EKS. When you configure the AWS check as a cluster check (required for an Agent running on Kubernetes), leave empty quotes for the parameters aws_access_key_id and aws_secret_access_key in the values.yaml file used to deploy the Cluster Agent.
If the Agent runs on an EC2 instance: The IAM role must have the following IAM policy. This policy grants the IAM principal permission to assume the role created in each target AWS account:
If StackState Agent does not run in an AWS environment, or if a role with the required policy is not available on the Agent's EC2 instance, an AWS user must be configured in the AWS check on the Agent.
Note that StackState also connects directly to AWS to pull CloudWatch metrics. StackState running within AWS can also be configured to authenticate using an IAM role, see IAM role for StackState on EC2 or EKS.
The StackState AWS CloudFormation Stack should be deployed in each AWS account that you will monitor. It provides the minimum level of access required for the StackState Agent to collect topology, telemetry and logs.
Quick deployment - Deploy all resources to a region in an account using a link.
StackState CloudFormation template - Download the StackState CloudFormation template to integrate into your own deployment workflow.
For special environments where the CloudFormation template may not function correctly, advanced AWS users can refer to the required AWS resources for a reference on all resources that must be manually created..
It is recommended to use the StackState CloudFormation template wherever possible as this provides an easy upgrade path for future versions and reduces the maintenance burden.
The necessary resources can be deployed for one account in a single region using an automated CloudFormation template.
The table below includes links to deploy the template in popular AWS regions. For any regions not listed, follow the steps described for the StackState template deployment.
To use OpenTelemetry Traces, set the IncludeOpenTelemetryTracing value to true.
You must be logged in to the target AWS account in the web console.
| Region Name | Template deployment link |
|---|---|
Ireland | |
Frankfurt | |
N. Virginia | |
Ohio | |
N. California | |
Hong Kong | |
Singapore | |
Sydney |
The default StackState CloudFormation template can be used to deploy all necessary resources. It can be deployed to multiple AWS accounts and regions at once by deploying it in a CloudFormation StackSet. It is recommended to use this template as it provides an easy upgrade path for future versions and reduces the maintenance burden compared to creating a custom template.
The template requires the following parameters:
MainRegion - The primary AWS region. This can be any region, as long as this region is the same for every template deployed within the AWS account. Global resources will be deployed in this region such as the IAM role and S3 bucket. Example: us-east-1.
StsAccountId - The 12-digit AWS account ID that is going to be monitored. This will be the AWS account that the IAM role can be assumed from, to perform actions on the target AWS account. Example: 0123456789012.
ExternalId - A shared secret that the StackState Agent will present when assuming a role. Use the same value across all AWS accounts that the Agent is monitoring. Example: uniquesecret!1.
IncludeOpenTelemetryTracing - Default: disabled. Set to enabled to include the OpenTelemetry layer in your deployment. This is required to retrieve OpenTelemetry traces from AWS Lambda scripts running NodeJS.
For more information on how to use StackSets, check the AWS documentation on working with AWS CloudFormation StackSets (docs.aws.amazon.com).
Install the AWS StackPack from the StackState UI StackPacks > Integrations screen. You will need to provide the following parameters, these will be used by StackState to query live telemetry from the AWS account. To create topology in StackState, you must configure the AWS check on StackState Agent V2.
Role ARN - the ARN of the IAM Role created by the cloudFormation stack. For example, arn:aws:iam::<account id>:role/StackStateAwsIntegrationRole where <account id> is the 12-digit AWS account ID that is being monitored.
External ID - a shared secret that StackState will present when assuming a role. Use the same value across all AWS accounts. For example, uniquesecret!1
AWS Access Key ID - The Access Key ID of the IAM user that will be used by StackState to collect CloudWatch metrics. This is the same as the IAM user used by the Agent to monitor AWS. If StackState is running within AWS, it is also possible authenticate with an IAM role.
AWS Secret Access Key - The Secret Access Key of the IAM user that will be used by StackState to collect CloudWatch metrics. This is the same as the IAM user used by the Agent to monitor AWS. If StackState is running within AWS, it is also possible to authenticate with an IAM role.
To enable the AWS check and begin collecting data from AWS, add the configuration below to StackState Agent V2.
If StackState Agent is running on Kubernetes, the AWS check should be configured as a cluster check.
If you do not already have it, you will need to add the StackState helm repository to the local helm client:
Update the values.yaml file used to deploy the cluster-agent with details of your AWS instance:
aws_access_key_id - The AWS Access Key ID. Leave empty quotes to use the IAM role attached to the EKS stackstate-cluster-agent pod.
aws_secret_access_key - The AWS Secret Access Key. Leave empty quotes to use the IAM role attached to the EKS stackstate-cluster-agent pod.
external_id - The same external ID used to create the CloudFormation stack in every account and region.
role_arn - In the example arn:aws:iam::123456789012:role/StackStateAwsIntegrationRole, substitute 123456789012 with the target AWS account ID to read.
regions - The Agent will only attempt to find resources in the specified regions. global is a special region for global resources, such as Route53.
Deploy the cluster_agent using the updated values.yaml:
Kubernetes:
OpenShift:
VPC FlowLogs can be analysed to retrieve relations between EC2 instances and RDS database instances. For each VPC that you want to analyse, a FlowLog needs to be configured. The process of adding FlowLogs for new VPCs could be automated using a Lambda triggered by a CloudTrail event that creates the FlowLog. Relations will be retrieved for EC2 instances and RDS database instances with a static public or private IP address and emit the proper URNs. For public IP addresses urn:host:/{ip-address}, for private IP addresses the URN has the form urn:vpcip:{vpc-id}/{ip-address}.
For further details, see Required AWS resources - VPC FlowLogs.
To configure a VPC FlowLog from the AWS console:
From the VPC Dashboard, choose Your VPCs under VIRTUAL PRIVATE CLOUD.
Select the VPC that you want to configure.
Select Flow logs on the lower TAB-bar.
Click Create flow log.
Add the settings as shown in the screenshot.
Access to the internet is required to call the AWS APIs. If StackState or the StackState Agent cannot be given direct internet access, an HTTP proxy can be used to proxy the API calls.
In the StackState AWS integration, CloudWatch metrics are pulled directly by StackState, while events and topology data are collected by the StackState Agent. This means that proxy details must be configured in two places to handle all requests - StackState for CloudWatch metrics and StackState Agent for topology and events data.
To configure a proxy for events and topology data collected by StackState Agent, see how to use an HTTP/HTTPS proxy for StackState Agent.
To configure a proxy for CloudWatch metrics collected by StackState, follow the steps below:
In the StackState UI, go to Settings > Telemetry Sources > CloudWatch sources.
Find the CloudWatch source for which you want to configure a proxy. Open the ... menu to the right and select Edit.
Enter the proxy details in Proxy URI.
Click TEST CONNECTION to check that the proxy can be used to successfully connect to CloudWatch.
Click UPDATE to save the proxy settings. Be aware that you may need to unlock it before this succeeds.
To check the status of the AWS integration, run the status subcommand and look for aws_topology under Running Checks:
The AWS integration retrieves the following data:
The AWS StackPack supports the following event:
EC2 Instance Run State: when the instance is started, stopped, or terminated. This will appear as the Run State in the EC2 instance component.
AWS events are primarily used to provide real-time updates to topology. These events are not displayed as StackState events.
Metrics data is pulled at a configured interval directly from AWS by the StackState CloudWatch plugin. Retrieved metrics are mapped onto the associated topology component.
The AWS service data shown below is available in StackState as components with the associated relations. The retrieved topology can be further enhanced by enabling OpenTelemetry traces.
| Service | Resource | Relations |
|---|---|---|
API Gateway | Method | SQS Queue, Lambda Function |
API Gateway | Method - HTTP Integration | |
API Gateway | Resource | API Gateway Method |
API Gateway | Rest API | API Gateway Stage |
API Gateway | Stage | API Gateway Resource |
Auto Scaling | Group | EC2 Instance, Classic Load Balancer, Auto Scaling Target Group |
CloudFormation | Stack | All Supported Resources*, Nested CloudFormation Stack |
DynamoDB | Stream | |
DynamoDB | Table | DynamoDB Stream |
EC2 | Instance | EC2 Security Group |
EC2 | Security Group | EC2 Instance |
EC2 | Subnet | EC2 Instance, EC2 VPC |
EC2 | VPC | EC2 Security Group, EC2 Subnet |
EC2 | VPN Gateway | EC2 VPC |
ECS | Cluster | EC2 Instance, ECS Service, ECS Task, Route53 Hosted Zone |
ECS | Service | Load Balancing Target Group, ECS Task |
ECS | Task | |
Kinesis | Data Stream | Kinesis Firehose Delivery Stream |
Kinesis | Firehose Delivery Stream | S3 Bucket |
Lambda | Alias | |
Lambda | Function | All Supported Resources* (Input), EC2 VPC, Lambda Alias, RDS Instance** |
Load Balancing | Application Load Balancer | EC2 VPC, Load Balancing Target Group, Load Balancing Target Group Instance |
Load Balancing | Classic Load Balancer | EC2 Instance, EC2 VPC |
Load Balancing | Network Load Balancer | EC2 VPC, Load Balancing Target Group, Load Balancing Target Group Instance |
Load Balancing | Target Group | EC2 VPC |
Load Balancing | Target Group Instance | EC2 Instance |
RDS | Cluster | RDS Instance |
RDS | Instance | EC2 VPC, EC2 Security Group |
Redshift | Cluster | EC2 VPC |
Route53 | Domain | |
Route53 | Hosted Zone | |
S3 | Bucket | Lambda Function |
SNS | Topic | All Supported Resources* |
SQS | Queue | |
Step Functions | Activity | |
Step Functions | State | Step Functions (All), Lambda Function, DynamoDB Table, SQS Queue, SNS Topic, ECS Cluster, Api Gateway Rest API |
Step Functions | State Machine | Step Functions (All) |
* "All Supported Resources" - relations will be made to any other resource on this list, should the resource type support it.
** This relation is made by finding valid URIs in the environment variables of the resource. For example, the DNS hostname of an RDS instance will create a relation.
OpenTelemetry creates traces from the AWS services that your Lambdas interacts with. Retrieved traces are available in the Traces Perspective and are also used to enhance the retrieved topology.
➡️ Learn more about how to set up and use OpenTelemetry
A high-level of overview of all resources necessary to run the StackState Agent with full capabilities is provided in the graph below. Users with intermediate to high level AWS skills can use these details to set up the StackState Agent resources manually. For the majority of installations, this is not the recommended approach. Use the provided StackState CloudFormation template unless there are environment-specific issues that must be worked around.
Hourly and event-based updates collect data:
Hourly full topology updates - collected by the StackState Agent using an IAM role with access to the AWS services.
Event-based updates for single components and relations - captured using AWS services and placed into an S3 bucket for the StackState Agent to read.
If the StackState Agent does not have permission to access a certain component, it will skip it.
The bare minimum necessary to run the StackState Agent is an IAM role with necessary permissions. The Agent will always attempt to fetch as much data as possible for the supported resources. If a permission is omitted, the Agent will attempt to create a component with the data it has.
For example, if the permission s3:GetBucketTagging is omitted, the Agent will fetch all S3 buckets and their associated configuration, but the tags section will be empty.
IAM is a global service. Only one IAM role is necessary per account.
Once the Agent has finished reading a file in this bucket, the file will be deleted. Do not use an existing bucket for this, the Agent should have its own bucket to read from. The S3 bucket will not be read from if it does not have bucket versioning enabled, to protect data.
The S3 bucket is used to store all incoming events from EventBridge and other event-based sources. The Agent then reads objects from this bucket. These events are used to provide features such as real-time topology updates, and creating relations between components based on event data such as VPC FlowLogs. If the S3 bucket is not available to the Agent it will fallback to reading CloudTrail directly, which introduces a 15 minute delay in real-time updates. EventBridge events and VPC FlowLogs are only available via the S3 bucket.
Only one S3 bucket is necessary per account; all regions can send to the same bucket.
A catch-all rule for listening to all events for services supported by the AWS StackPack. All matched rules are sent to a Kinesis Firehose delivery stream.
EventBridge IAM Role - JSON - Give permission for EventBridge to send data to Kinesis Firehose
A rule must be created in each region where events are captured, each sending to a Firehose delivery stream in the same region.
Kinesis Firehose is used to receive and batch events from EventBridge. This delivery stream batches events per 60 seconds and pushes an object to S3. 60 seconds is the recommended value - setting this value any higher will negligibly decrease storage costs while increasing the delay in topology updates.
The Prefix must be set to AWSLogs/${AccountId}/EventBridge/${Region}/, where ${AccountId} and ${Region} are the account ID and region, for example, eu-west-1. Files must be compressed using the GZIP option.
A delivery stream must be created in each region where events are captured, however, the target S3 bucket can exist in any region.
Kinesis Firehose IAM Role - JSON - Gives permission for Firehose to send data to an S3 bucket.
A KMS Customer Managed Key (CMK) can be used to secure data at rest in S3. The KMS key is used in the Firehose Delivery Stream. The S3 bucket also uses the KMS key as its default key.
Use of a KMS is key is not necessary for the operation of the StackPack, however as encryption at rest is a requirement in most environments, the CloudFormation template includes this by default.
A KMS key must be created in each region where events are captured.
VPC FlowLogs support is currently experimental.
A VPC configured to send flow logs to the stackstate-logs-${AccountId} S3 bucket. The Agent requires the AWS default format for VPC FlowLogs, and expects data to be aggregated every 1 minute. The FlowLogs contain meta information about the network traffic inside VPCs. Only private network traffic is considered, traffic from NAT gateways and application load balancers will be ignored.
S3 objects that have been processed will be deleted from the bucket to make sure they will not be processed again. On the default S3 bucket, object versioning is enabled, this means objects will not actually be immediately deleted. A lifecycle configuration will expire (delete) both current and non-current object versions after one day. When using a non default bucket, you can set these expiry periods differently.
If configuring FlowLogs using CloudFormation, the stackstate-resources template exports the ARN of the S3 bucket it creates, so this can be imported into your template.
A FlowLog must be configured for each VPC that you want to analyse.
The AWS StackPack CloudFormation template contains all resources that are necessary to run the AWS check on the StackState Agent. The installed resources are kept as minimal as possible. All costs incurred are minimal but variable, with costs scaling depending on how many events are emitted in a given account. In practice, the costs created by the AWS integration will be negligible.
Kinesis Firehose: priced by the amount of data processed. Events use very small amounts of data. Firehose pricing (aws.amazon.com)
S3: priced by amount of data stored, and amount of data transferred. Running the Agent inside of AWS will reduce data transfer costs. S3 pricing (aws.amazon.com)
KMS: a flat fee of $1 per month per key, with additional costs per request. KMS pricing (aws.amazon.com)
CloudWatch metrics: priced per metric retrived. Metrics are only retrieved when viewed or when a check is configured on a CloudWatch metric. CloudWatch pricing (aws.amazon.com)
When the AWS integration is enabled, three views will be created in StackState for each instance of the StackPack.
AWS - [instance_name] - All - includes all resources retrieved from AWS by the StackPack instance.
AWS - [instance_name] - Infrastructure - includes only Networking, Storage and Machines resources retrieved from AWS by the StackPack instance.
AWS - [instance_name] - Serverless - includes only S3 buckets, lambdas and application load balancers retrieved from AWS by the StackPack instance.
Components retrieved from AWS will have an additional action available in the component context menu and component details pane on the right-hand side of the screen. This provides a deep link through to the relevant AWS console at the correct point.
For example, in the StackState Topology Perspective:
Components of type aws-subnet have the action Go to Subnet console, which links directly to this component in the AWS Subnet console.
Components of type ec2-instance have the action Go to EC2 console, which links directly to this component in the EC2 console.
On import, all topology in StackState will be given the label stackpack:aws-v2.
Topology imported by the AWS (Legacy) integration will have the label stackpack:aws.
Any tags that exist in AWS will be added to the StackState topology as labels. In addition, the following special tags can be added in AWS to influence how the topology is built in StackState:
| Tag | Description |
|---|---|
| Adds the specified value as an identifier to the StackState component |
| Places the StackState component in the environment specified |
Check the StackState support site for:
To uninstall the StackState AWS StackPack, click the Uninstall button from the StackState UI StackPacks > Integrations > AWS screen. This will remove all AWS specific configuration in StackState.
Once the AWS StackPack has been uninstalled, you will need to delete the StackState AWS Cloudformation stack from the AWS account being monitored. This can be done using the web console or the command line.
To delete the StackState AWS Cloudformation stack from an AWS account using the web console: If the template is in the main region, the S3 bucket used by StackState must be emptied as CloudFormation can't delete an empty bucket. Follow these steps:
Disable the EventBridge rule. Go to EventBridge, and find the rule name starting with stackstate-resources-StsEventBridgeRule. Open this rule, and press the "Disable" button.
Delete all flowlogs that send to this bucket. Go to the VPC service, and select each VPC in the VPCs list. Look in the FlowLogs tab in the details section, and delete any flowlogs that are sent to the S3 bucket starting with stackstate-logs.
Go to the S3 service. Select (don't open) the bucket named stackstate-logs-${AccountId} where ${AccountId} is the 12-digit identifier of your AWS account.
Select "Empty", and follow the steps to delete all objects in the bucket.
Go to the CloudFormation service. Select the CloudFormation template. This will be named stackstate-resources if created via the quick deploy method, otherwise the name was user-defined.
In the top right of the console, select "Delete".
These steps assume you already have the AWS CLI installed and configured with access to the target account. If not, follow the AWS documentation.
If --region is the main region, follow these steps to delete the S3 bucket. Before emptying the bucket, disable any event sources that are sending files to the bucket. This is a versioned S3 bucket, so each object version must be deleted individually. If there are more than 1000 items in the bucket this command will fail; it's likely more convenient to perform this in the web console.
Delete the CloudFormation template: aws cloudformation delete-stack --stack-name stackstate-resources --region <region>.
Find out how to uninstall using a specific AWS profile or an IAM role (docs.aws.amazon.com).
AWS StackPack v1.2.0 (2022-03-03)
Improvement: Added OpenTelemetry information STAC-15902
AWS StackPack v1.1.4 (2021-11-16)
Improvement: Updated AWS CLI prerequisite text
AWS StackPack v1.1.3 (2021-11-15)
Improvement: Documentation update - Added AWS CLI as a prerequisite to set up the integration
AWS StackPack v1.1.2 (2021-11-04)
Improvement: Documentation update - improved installation instructions
AWS StackPack v1.1.1 (2021-10-25)
Improvement: Documentation update regarding experimental nature of Flow Logs support
AWS StackPack v1.1.0 (2021-10-14)
Improvement: Flow Logs support (experimental)
AWS StackPack v1.0.3 (2021-09-23)
Bugfix: Fixed CloudWatch telemetry source configuration for use-role StackPack install option
AWS StackPack v1.0.2 (2021-08-20)
Improvement: Add description to Views
AWS StackPack v1.0.1 (2021-07-23)
Bugfix: Use proper domain
