Monitoring Lambda Managed Instances
LMI publishes CloudWatch metrics so you can watch fleet utilisation, remaining capacity, and per-function execution environment behaviour. Metrics help right-size memory, interpret scaling, and spot concurrency pressure.
For full metric lists, dimensions, and console navigation, use Monitoring Lambda Managed Instances as the product reference.
Two metric scopes
Section titled “Two metric scopes”| Scope | What it answers | Example dimensions (per AWS) |
|---|---|---|
| Capacity provider | How hard the managed fleet is working and how much headroom remains | CapacityProviderName, InstanceType |
| Execution environment | How a specific function version uses CPU, memory, and concurrency on that fleet | CapacityProviderName, FunctionName, Resource |
For LMI, the Resource dimension on execution-environment metrics targets function versions — format <FunctionName>:<FunctionVersion> (not aliases-only semantics in the same way).
Highlights (what to watch)
Section titled “Highlights (what to watch)”Capacity provider (fleet) level
Section titled “Capacity provider (fleet) level”| Metric (names per AWS) | Use |
|---|---|
| CPUUtilization, MemoryUtilization | Overall pressure across instances behind the provider |
| vCPUAvailable, MemoryAvailable | Headroom still allocatable to environments |
| vCPUAllocated, MemoryAllocated | What is already committed to execution environments |
Execution environment (function version) level
Section titled “Execution environment (function version) level”| Metric (names per AWS) | Use |
|---|---|
| ExecutionEnvironmentConcurrency | Observed concurrency (sampled over a window per AWS) |
| ExecutionEnvironmentConcurrencyLimit | Ceiling per environment you configured |
| ExecutionEnvironmentCPUUtilization, ExecutionEnvironmentMemoryUtilization | Whether that version is sized correctly |
Cadence and retention
Section titled “Cadence and retention”Per AWS, LMI metrics are published on 5-minute intervals and retained 15 months — plan dashboards and alarms around that granularity.
See also
Section titled “See also”Elasticity & CPUHow CPU-linked scaling relates to the fleet metrics above.
Concurrency & runtimesWhat execution environments are before you interpret concurrency metrics.
QuotasService and throughput limits that interact with observed load.