AI agent skill · by @automateyournetwork

Gcp Cloud Monitoring

Google Cloud Monitoring — time series metrics, alert policies, active alerts, metric discovery. Use when checking GCP network performance, investigating firing alerts, querying VM CPU or memory metrics, reviewing Cloud VPN tunnel status, or assessing load balancer latency.

Archived 2026-10-23 · Source updated 2026-10-11 · License: Apache-2.0. Check the original repository for the latest version.

SKILL.md instructions

# GCP Cloud Monitoring

## MCP Server

- **Endpoint**: `https://monitoring.googleapis.com/mcp` (Streamable HTTP)
- **Auth**: OAuth 2.0 via Google IAM — service account key (`GOOGLE_APPLICATION_CREDENTIALS`) or `gcloud auth application-default login`
- **Requires**: `GCP_PROJECT_ID` environment variable

## Available Tools (6)

| Tool | What It Does |
|------|-------------|
| `list_timeseries` | Query time series data — CPU, memory, network, disk metrics for any GCP resource |
| `list_metric_descriptors` | Discover available metric types in a project — find what you can monitor |
| `list_alert_policies` | List all alerting policies — conditions, notification channels, thresholds |
| `get_alert_policy` | Get details of a specific alerting policy |
| `list_alerts` | List current and past alert violations — what's firing right now |
| `get_alert` | Get details of a specific alert violation |

## Workflow: GCP Network Monitoring

When a user asks "how is our GCP network performing?":

1. **Check alerts**: `list_alerts` to find any active alert violations
2. **VM network metrics**: `list_timeseries` for `compute.googleapis.com/instance/network/received_bytes_count` and `sent_bytes_count`
3. **Packet drops**: `list_timeseries` for `compute.googleapis.com/instance/network/received_packets_dropped_count`
4. **Firewall metrics**: `list_timeseries` for `compute.googleapis.com/firewall/dropped_packets_count`
5. **Load balancer metrics**: `list_timeseries` for `loadbalancing.googleapis.com/https/request_count` and `total_latencies`
6. **Report**: Network health dashboard with any issues flagged

## Workflow: Alert Investigation

When investigating GCP alerts:

1. **List active alerts**: `list_alerts` — find what's currently firing
2. **Get alert details**: `get_alert` — condition, threshold, resource affected
3. **Get policy**: `get_alert_policy` — what triggers this alert, notification channels
4. **Pull metrics**: `list_timeseries` for the affected metric — see the spike/anomaly
5. **Cross-reference**: Use `gcp-cloud-logging` for correlated log entries
6. **Report**: Alert investigation with root cause and timeline

## Workflow: Resource Health Check

When checking GCP infrastructure health:

1. **Discover metrics**: `list_metric_descriptors` filtered by service (compute, networking, loadbalancing)
2. **VM CPU/Memory**: `list_timeseries` for `compute.googleapis.com/instance/cpu/utilization` and memory metrics
3. **Disk I/O**: `list_timeseries` for `compute.googleapis.com/instance/disk/read_bytes_count` and write metrics
4. **Network throughput**: `list_timeseries` for network sent/received bytes
5. **Alert status**: `list_alert_policies` + `list_alerts` — any policies in violation?
6. **Report**: Infrastructure health dashboard with severity ratings

## Common GCP Network Metrics

| Metric | What It Tells You |
|--------|-------------------|
| `compute.googleapis.com/instance/network/received_bytes_count` | Inbound network throughput per VM |
| `compute.googleapis.com/instance/network/sent_bytes_count` | Outbound network throughput per VM |
| `compute.googleapis.com/instance/network/received_packets_dropped_count` | Dropped inbound packets (congestion) |
| `compute.googleapis.com/instance/network/sent_packets_dropped_count` | Dropped outbound packets (congestion) |
| `compute.googleapis.com/firewall/dropped_packets_count` | Packets dropped by VPC firewall rules |
| `loadbalancing.googleapis.com/https/request_count` | HTTP(S) LB request rate |
| `loadbalancing.googleapis.com/https/total_latencies` | HTTP(S) LB end-to-end latency |
| `loadbalancing.googleapis.com/https/backend_latencies` | Backend response time behind LB |
| `vpn.googleapis.com/tunnel_established` | Cloud VPN tunnel state (1=up, 0=down) |
| `vpn.googleapis.com/sent_bytes_count` | Bytes sent through VPN tunnel |
| `router.googleapis.com/bgp/received_routes_count` | BGP routes received by Cloud Router |
| `interconnect.googleapis.com/link/received_bytes_count` | Cloud Interconnect link throughput |

## Important Rules

- **Remote MCP server** — hosted by Google, no local install needed
- **OAuth 2.0 authentication** — uses IAM for access control
- **Project-scoped** — metrics are scoped to the configured GCP project
- **Read-only** — monitoring queries don't modify anything
- **Record in GAIT** — log monitoring investigations for audit trail

## Environment Variables

- `GCP_PROJECT_ID` — Google Cloud project ID
- `GOOGLE_APPLICATION_CREDENTIALS` — Path to service account key JSON file

## Failure Behavior

- On a tool error (timeout, unreachable host, malformed response), report the failure and its error message directly to the user rather than fabricating or guessing at results.
- For a confirmed read-only call, check connectivity and retry once if appropriate. For any call that changes state or sends a message, a timeout does not prove the action failed: inspect current state or delivery status before retrying, preserve the required approval/change gates, and do not repeat an action whose outcome is unknown.