Blast Radius Analysis & Subgraph Isolation
Transitive BFS impact graph traversal revealing every resource placed at risk before running apply.
Blast Radius HUD Pill
On-canvas floating HUD breaking down direct casualties, transitive casualties, and destructive replacements in real time.
Subgraph Isolation Mode
Dedicated toggle ([B]) that dims or removes all unrelated infrastructure to focus purely on the impacted dependency cascade.
Depth Filter Stepper
Interactive stepper switching between 1-hop direct dependencies, 2-hop intermediate links, and full transitive closure.
Mutating Only Filter
Filters out read-only and no-op downstream observers, isolating only resources undergoing destructive or stateful alterations.
The Problem
Hidden Cascading Re-creations
In Terraform, changing an attribute on a low-level primitive (such as an AWS Subnet CIDR, Security Group, or KMS Key) often forces replacement (`+/-`). Because dependent services bind to the resource's physical ID, the replacement cascades downstream, tearing down databases and Kubernetes node groups without warning.
Under the Hood: Engine & Algorithm
Bidirectional Breadth-First Traversal
When a node is selected, plan-parse conducts a bidirectional BFS traversal along both explicit `depends_on` arcs and implicit HCL expression references. It indexes parents and children, classifies casualties by severity, and animates high-voltage signal paths across the DAG.
Workstation Flow: How to Use
Click a node on the canvas to immediately activate the Blast Radius HUD and signal paths.
Filter canvas rendering down to the exact subgraph of casualties affected by the selected node.
Step through 1-hop, 2-hop, or full transitive depth to inspect immediate vs distant ripple effects.
Real-World SRE & Platform Scenario
Production Outage Prevention
An engineer refactors a shared database subnet group name. Standard terminal output showed 1 replacement. Blast Radius reveals 2 production RDS clusters, 4 read replicas, and 12 ECS tasks would be destroyed and recreated during deployment.