GeassNetworkLog

A GeassNetworkLog is a schema for one HTTP access-log record: the project, environment and service, the host, method and path of a request, the status code and the latency. It holds request metadata only, never request or response bodies.

Caution

Nothing in Geass creates these objects, and nothing reads them. The operator has a controller that deletes records older than seven days, and the platform configuration has a traefikAccessLogs block, but no code turns Traefik’s access log into GeassNetworkLog records, and the dashboard has no screen for them. The kind is a stub for a feature that is not built. For HTTP request data today, use the request metrics on a service’s Metrics tab (see Observability).

Summary

Property	Value
API version	geass.geass.dev/v1alpha1
Kind	GeassNetworkLog
Plural	geassnetworklogs
Scope	Namespaced
Namespace	geass-system, by convention
Subresources	None. The kind has no status.
Printer columns	None

Example

This is what a record looks like if something writes one.

storefront-request.yaml

apiVersion: geass.geass.dev/v1alpha1
kind: GeassNetworkLog
metadata:
  name: storefront-1760000000-a1b2c3
  namespace: geass-system
spec:
  project: proj-1a2b3c4d
  environment: production
  app: storefront
  timestamp: "2026-10-12T09:00:00Z"
  host: shop.example.com
  method: GET
  path: /checkout
  status: 200
  latencyMillis: 42
  upstreamOutcome: ok

Spec fields

Field	Type	Default	Validation	Meaning
spec.project	string	none	Required	The project ID.
spec.environment	string	none	Required	The environment.
spec.app	string	none	Required	The service that answered.
spec.timestamp	time	none	Required	When the request happened. The retention check uses it.
spec.host	string	none	none	The Host of the request.
spec.method	string	none	none	The HTTP method.
spec.path	string	none	none	The request path.
spec.status	integer	none	32-bit integer	The response status code.
spec.latencyMillis	integer	none	64-bit integer	How long the request took, in milliseconds.
spec.upstreamOutcome	string	none	none	A free-form note on how the upstream answered.

What happens under the hood

The operator runs a controller for the kind. For each record it compares spec.timestamp with the retention, 7 days by default, and deletes the record when it is older. If it is not old enough, it looks at the record again after half the retention (3.5 days), so a record can outlive its retention by up to that long. A record with an empty timestamp is never deleted. The operator’s role can create, delete, get, list and watch the kind, and it cannot update it.

The retention is a field of the controller, and the operator does not set it from any flag, environment variable or resource, so 7 days is what you get.

The traefikAccessLogs block of the GeassPlatformConfig (enabled, format, networkLogEndpoint) is in the schema and nothing reads it.

Look for records

kubectl -n geass-system get geassnetworklogs

On a cluster installed from the current code this lists nothing.

Editing by hand

You can create and delete records, but there is nothing to gain: they do not appear anywhere, and the operator removes them after seven days. Do not store personal data in them.

Troubleshooting

The list is empty. That is expected. Nothing writes these records.

I set traefikAccessLogs.enabled on the platform configuration and nothing happens. The field is not used. See the caution at the top of this page.

Related

• Observability
• GeassPlatformConfig
• Resources overview