Metrics & Dashboards

Metrics are the heartbeat of your monitoring strategy. They are numerical measurements collected at regular intervals that describe the behaviour of your systems over time. Google Cloud Monitoring provides a powerful metrics infrastructure with automatic collection, flexible querying, and rich visualisation through dashboards.

---

Understanding GCP Metrics

Metric Kinds

Every metric in Cloud Monitoring has a kind that describes how values relate over time:

Kind	Description	Example
Gauge	A value at a specific point in time	CPU utilisation (45%), memory usage (2.3 GiB)
Delta	The change in value over a time interval	Request count in the last minute, bytes sent in the last 5 minutes
Cumulative	A running total since a fixed start time	Total requests since process start, total bytes transferred

Understanding metric kinds is critical when choosing aggregation functions. You cannot meaningfully sum gauge metrics across time — you would use mean or max instead.

Value Types

Metrics also have a value type that determines the data format:

Value Type	Description
INT64	A 64-bit integer (e.g., request count)
DOUBLE	A 64-bit floating-point number (e.g., CPU utilisation as a fraction)
BOOL	True or false (e.g., instance is running)
STRING	A text value (rare — used for status labels)
DISTRIBUTION	A histogram of values (e.g., latency distribution)

---

Metrics Explorer

Metrics Explorer is the primary tool for ad-hoc metric analysis in the Cloud Console. It provides a flexible query interface for exploring any metric across your monitored resources.

Building a Query

1. Select a resource type — e.g., VM Instance, Cloud Run Revision, Cloud SQL Database
2. Select a metric — e.g., CPU utilisation, request count, disk read bytes
3. Add filters — narrow results by labels such as zone, instance name, or project
4. Set aggregation — choose how to combine multiple time series (sum, mean, max, percentile)
5. Set alignment — define the time window for each data point (1 minute, 5 minutes, 1 hour)

Aggregation and Alignment

When you have multiple time series (e.g., CPU utilisation for 50 VMs), you need to aggregate them into a meaningful view:

Aggregation	Use Case
Sum	Total request count across all instances
Mean	Average CPU utilisation across a fleet
Max	Peak memory usage across all instances
99th percentile	Tail latency for a service
Count	Number of time series matching a filter

Alignment divides the time axis into regular intervals and produces one value per interval. A 5-minute alignment with mean aligner produces the average value for each 5-minute window.

---

Monitoring Query Language (MQL)

For advanced queries, Cloud Monitoring supports MQL — a text-based query language that provides more flexibility than the visual query builder:

fetch gce_instance
  | metric 'compute.googleapis.com/instance/cpu/utilization'
  | filter zone = 'europe-west2-a'
  | group_by [instance_name], mean(val())
  | align rate(5m)
  | every 5m

MQL Advantages

• Arithmetic operations — combine metrics (e.g., calculate cache hit ratio)
• Joins — correlate metrics from different resource types
• Conditional logic — apply thresholds and filters programmatically
• Reusability — save and share complex queries as text

Common MQL Patterns

# Calculate error rate as a percentage
fetch cloud_run_revision
  | metric 'run.googleapis.com/request_count'
  | filter response_code_class != '2xx'
  | group_by [], sum(val())
  | div
    (fetch cloud_run_revision
     | metric 'run.googleapis.com/request_count'
     | group_by [], sum(val()))
  | mul 100

---

Dashboards