Skill 121 · Setting Up CloudWatch Observability
Subchapter 121.10
references/cloudwatch-omni/gcp-ingestion/gcp-ingestion.mdMarkdown2 KBView on GitHub
CloudWatch Application Observability ingests from Google Cloud — the logs, metrics, and traces their service emits — using the Amazon CloudWatch agent.
Decide by the customer’s intent, not by whether they already name a mechanism. Most customers won’t know the agent or any setup process — guiding them there is the whole point — so routing on keywords like “CloudWatch agent” is wrong. Route on what telemetry they are trying to collect.
Ask what the customer actually wants to ingest:
The telemetry their own application/service produces — the logs they write
in their own code, and the metrics and traces their application emits. This is
the supported path: load custom-telemetry.md and guide them through
installing the CloudWatch agent on their GCE instance or GKE cluster.
Telemetry their Google Cloud resources emit on their own — resource and platform logs and metrics they did not author, the Google Cloud equivalent of AWS VPC flow logs or Route 53 query logs (Cloud Audit Logs and the VPC flow logs Google Cloud emits for them). This is not supported. Tell the customer this directly and stop. Do not offer the CloudWatch agent as an alternative or workaround for this: the agent only ships the customer’s own application telemetry and cannot collect telemetry that Google Cloud resources emit on their own, so suggesting it here would be misleading regardless of what else the feature list gains.
If the intent is unclear, ask exactly one question to disambiguate:
Are you trying to send the telemetry your own application produces (the logs/metrics/traces from your code), or telemetry that your Google Cloud resources emit on their own (like Cloud Audit Logs)?
The former is supported; the latter is not.
For the customer’s own application telemetry, follow custom-telemetry.md.