metric_filters | Patterned Designs Documentation

Overview

Metric filters define how CloudWatch Logs transforms log data into CloudWatch metrics. Use metric filters to extract metric observations from ingested log events and create alarms.

Filter Object Structure

metric_filters = [
  {
    name             = "error-count"
    pattern          = "ERROR"
    metric_name      = "ErrorCount"
    metric_namespace = "MyApp"
    metric_value     = "1"
  }
]

Field	Required	Description
name	Yes	Unique name for the filter
pattern	Yes	Filter pattern to match log events
metric_name	Yes	Name of the CloudWatch metric
metric_namespace	Yes	Namespace for the metric
metric_value	No	Value to publish (default: “1”)

Common Patterns

Error Counting

{
  name             = "errors"
  pattern          = "[timestamp, level=ERROR, ...]"
  metric_name      = "ErrorCount"
  metric_namespace = "Application"
}

Latency Extraction

{
  name             = "latency"
  pattern          = "[timestamp, level, message, latency]"
  metric_name      = "RequestLatency"
  metric_namespace = "Application"
  metric_value     = "$latency"
}

Best Practices

Use descriptive filter names
Group related metrics in the same namespace
Create alarms on critical metrics
Test patterns with sample log data first

Full Module Example

module "cloudwatch_logs" {
  source  = "registry.patterneddesigns.ca/essentials/cloudwatch-logs/aws"
  version = "1.0.0"

  # metric_filters
  metric_filters = [{name = "ErrorCount", pattern = "ERROR", metric_name = "ErrorCount", metric_namespace = "MyApp"}]

  # Other required inputs
  log_group_name = "/aws/lambda/my-function"
}