tier), instance IDs, or VPC, and group instances into application tiers (such as web or worker). This helps answer questions like “which application tier(s) plausibly drive load on this RDS?” when investigating non-Kubernetes alerts or bridging load balancers (ELB) and databases (RDS).
All EC2 API calls are read-only and routed through the shared aws_sdk_client allowlist, so the integration cannot mutate your EC2 resources.
Prerequisites
- AWS credentials configured per the AWS integration (role ARN recommended)
- An EC2 environment with tagged instances, explicit instance IDs, or a VPC you want OpenSRE to inspect
- IAM permissions for
ec2:DescribeInstances
Setup
Environment variables
The tool accepts parameter hints from the resolved
ec2 source:
tiers: List of tier names in the resolvedec2source (e.g.["web"]or["web", "worker"]). The parameter extractor maps a single item to the singulartiertool argument, or leavestierempty when multiple tiers are in scope so instances across all tiers are enumerated and returned grouped inby_tier.instance_ids: List of explicit EC2 instance IDs.vpc_id: VPC ID filter to limit discovery to a specific network.region: AWS region override (defaults toAWS_REGIONorus-east-1).
IAM permissions
The integration only needs read-only EC2 describe permissions:ReadOnlyAccess policy, ec2:DescribeInstances is already included.
Verify
ec2 verify target. A successful AWS verify confirms your AWS credentials are valid; the EC2 tool becomes available whenever tiers, instance_ids, vpc_id, or _backend is present in the resolved ec2 source.
Expected output:
Tools
The tool becomes available to the planner whenever
tiers, instance_ids, vpc_id, or _backend is present in the resolved ec2 sources.
Use cases
- Discovering EC2 application tiers when investigating a non-Kubernetes alert
- Mapping a tier name (
web,worker, etc.) to its active instance IDs - Bridging EC2 → RDS when answering “which application tier drives database load”
Gotcha
Terminated or stopped EC2 instances (terminated, stopped, stopping, shutting-down) are automatically excluded from the tool’s results so they do not distort load-attribution or topology analysis. If an expected instance does not appear in the returned list, verify that the instance is in the running state and properly tagged with tier or Tier. Additionally, at least one filter (tier, instance_ids, or vpc_id) must be provided in the tool call or resolved sources, or the tool will return an unavailable status.