Alert and Provider have no status conditions — cannot see last dispatch or last error

[RaviTharuma]

Describe the bug / problem

Alert and Provider objects expose no status conditions (and essentially no status at all) that operators can use to answer:

• Was the last event delivered?
• When was the last successful dispatch?
• What was the last provider error (401, 429, network, etc.)?

Live observation on notification-controller v1.9.2 (API notification.toolkit.fluxcd.io/v1beta3):

• Alert.status is completely empty: status: {} (no keys).
• Provider.status is empty the same way.
• Events are sometimes delivered to the chat provider, so the pipeline is not dead — but the CRs themselves give zero operational signal.
• kubectl get alert effectively shows Age only — no Ready column, no lastSentAt, no lastError.

Debugging provider failures (Telegram 401 auth, 429 rate limit, DNS flaps, etc.) requires scraping notification-controller logs. That is a poor operator experience for a component whose whole job is reliability of outbound notifications.

This is distinct from #1370 (flood volume / transient DNS cascades / exclusionList). That issue is about how many events fire. This issue is about observability of the Alert and Provider objects themselves.

Steps to reproduce

1. Configure a Provider (e.g. type: telegram) and an Alert v1beta3 with eventSeverity: error and broad eventSources (GitRepository / Kustomization / HelmRelease / HelmRepository / OCIRepository / ImageUpdateAutomation wildcards are enough to generate traffic).
2. Confirm some events reach the provider (or intentionally break credentials / hit rate limits).
3. Inspect the CRs:

kubectl get alert -A
# Age only; no Ready / last error columns

kubectl get alert <name> -n <ns> -o yaml
# status: {}

kubectl get provider <name> -n <ns> -o yaml
# status empty / no conditions

4. kubectl describe alert does not surface last Telegram (or other provider) error either — logs remain the only source of truth.

Expected behavior

Status that answers “is notification path healthy?” without log diving, for example:

Field / condition	Purpose
status.conditions Ready	Provider config valid; controller can authenticate / reach endpoint
lastHandledEvent (or similar)	Identity of last event considered (name/kind/message hash)
lastDispatchTime	When the last successful send completed
lastDispatchError	Last failure message (redacted — no tokens, no full chat IDs, no secrets)

Printer columns for Ready + last error age would make kubectl get alert,provider useful for triage.

Even a minimal Ready condition (config OK vs last dispatch failed) would be a large step up from status: {}.

Environment

• Kubernetes: single-node (generic)
• Flux CLI: v2.9.4
• Distribution: flux-v2.9.1
• notification-controller: v1.9.2
• Alert API: v1beta3, eventSeverity: error
• Provider: telegram (channel ID omitted)
• Alert.spec.exclusionList already populated with generic regexes for connection refused / dial tcp lookup / empty tag DB / scan failed patterns — still no status reflecting exclusions vs sends

Additional context / non-goals

• Not asking to change event volume defaults here (see Reduce Telegram/provider alert floods during transient DNS/registry outages #1370).
• Not asking for full event history in etcd — counters / last-N summary is enough.
• Related: without status, operators cannot see how many events were excluded by exclusionList vs actually sent — that metric would also help tune regexes safely.

Code of Conduct

• I agree to follow this project's Code of Conduct

[RaviTharuma]

Related:

• Flood volume / transient cascades: Reduce Telegram/provider alert floods during transient DNS/registry outages #1370
• Docs: Alert/Provider lack of status noted in docs: OCI HelmRepository status is empty by design — document how to verify website#2643