Niklas Ye
bde5611b16
terdut-server v0.49.0 graphs the metric behind each alert on the incident page, but only queries hosts listed in TERDUT_PROMETHEUS_ALLOWED_HOSTS, because the URL comes out of a webhook payload. An install this operator creates had no way to set it short of spec.pod.extraEnv. spec.prometheus takes allowedHosts, preroll and timeout and maps them to TERDUT_PROMETHEUS_ALLOWED_HOSTS, _PREROLL and _TIMEOUT. Nothing is emitted unless allowedHosts is non-empty, so existing TerdutServers get the same Deployment as before, and unset durations are left to the server's defaults. There is no CRD default for the host: which Prometheus to trust is the deployer's call, not something this operator should assume. The CRD in config/ is regenerated and the chart's copy carries the same schema; the chart's optional TerdutServer passes the block through. Claude-Session: https://claude.ai/code/session_016mBLURvJoMuUEr9cB2RpUN
Terdut operator
Exposes terdut-server's configuration as Kubernetes objects, so teams can self-service it over gitops. See DESIGN.md for the design (read its revision section first: it is the current shape of credentials and the CRD catalog), and examples/demo for a working install.
CRDs
TerdutServer
A terdut-server install: Deployment, Service, database wiring (a DSN, or a Zalando
postgresClusterRef), and an operator key Secret it hands to the server so the operator can
authenticate. allowedTeams consents to TerdutTeams in other namespaces.
TerdutTeam
One team on a server, possibly in another namespace (serverRef, gated by that server's
allowedTeams). It carries the team's whole configuration:
displayNameand OIDC group bindingsescalation— the escalation ladderdeadmanSwitches— dead man's switches, by name (switches not listed are removed)
TerdutAlertSource
An alert-ingest integration on a team (teamRef). The server shows the webhook key once; it is
surfaced only through a generated Secret next to the object, never set explicitly.
Demo
examples/demo wires one of each together — two teams, each with an
escalation ladder, a dead man's switch and an alert source — plus a script that fires synthetic
Alertmanager webhooks at it, so you can watch incidents open, escalate and resolve without a real
Alertmanager.