Adds MQTT broker and Redis connection env vars to the fiesta
StatefulSet. Also corrects the image tag, which had drifted stale
in the repo at v1.3.90 while the cluster and the server copy of
the manifest were both already running v1.3.93.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The "gateway" sidecar in the fiesta StatefulSet (workolik360/alaska image,
running fiesta-gateway-script) and the k8s routing objects meant to reach it
(fiesta-route HTTPRoute, the port-8000 rules in nearle-ingress) never
actually served real traffic. fiesta.nearle.app is routed by a standalone
Docker/nginx proxy (conf/nginx-fiesta.conf, unchanged since the initial
commit) straight to NodePort 30823 - the real backend, no path splitting.
The Kubernetes-native routing objects have no working controller on this
cluster (no Traefik/Envoy pod; the shared Gateway resource's own status is
"Waiting for controller"), so they were inert either way.
The sidecar's only real effect, when briefly reachable, was publishing
Fiesta orders onto the same NATS subject worker-orders drains - which
forwards to jupiter, not this backend - producing the header-only phantom
orders fixed in the app on 2026-07-29. It's not a working parallel path,
it was the source of that bug.
Removed: the gateway container + gateway-script volume from the fiesta
StatefulSet, port 8000 from the fiesta Service, the fiesta-route HTTPRoute,
the fiesta-gateway-script ConfigMap (fiesta-gateway.yaml deleted entirely),
and the two dead port-8000 path rules in nearle-ingress's fiesta host block
(kept the correct catch-all). queue.workolik.com and the core-namespace NATS
worker pipeline are untouched - separate system entirely.
Verified live: fiesta pods rolled to single-container, service has only
port 80, and a real request through fiesta.nearle.app still behaves
correctly post-cleanup.
These two paths were routed through the NATS gateway sidecar (port 8000,
fiesta-gateway-script), whose only NATS consumer (worker-orders/products in
core) has EXTERNAL_BASE_URL hardcoded to jupiter. Fiesta has no NATS
consumer of its own, so every mobile order/product submitted through this
route was silently created on jupiter (header-only, wrong schema) instead
of ever reaching this backend's CreateOrderv3 - no orderdetails, no stock
movement, no error surfaced to the client (fire-and-forget "accepted").
Both paths are already registered directly on this backend, so they now
fall through the catch-all rule straight to it (port 80), synchronously.
Verified live: a real order now moves stock and a no-items order gets an
immediate 400 instead of a silent phantom accept.
Also corrects the image tag (v1.3.78 -> v1.3.90) to match what's actually
running - it had drifted since Flux was removed and cluster changes now
happen via direct kubectl.
- Rebuild manifests/doormile/miletruth.yaml (was corrupted since the
initial commit - contained pasted AI/terminal output, truncated env
var names/values, duplicate keys). Rebuilt from the confirmed-live
config, secrets sourced via a Secret instead of plaintext values.
- Lock down the Kubernetes Dashboard: remove --enable-skip-login /
--enable-insecure-login / --insecure-port=9090, remove the extra
cluster-admin binding on the dashboard's own ServiceAccount, remove
the now-dead insecure NodePort Service. Token-based login via the
existing admin-user ServiceAccount is unaffected.
- Fix the duplicate `backendRefs` key under the same HTTPRoute rule in
alaska.yaml (invalid/redundant YAML).
- Delete 6 redundant duplicate manifests (fiesta-sts/svc,
atlantis-sts/svc, jupiter-sts/svc) that were partial, stale subsets
of nearle-fiesta/atlantis/jupiter.yaml - one pair disagreed on the
fiesta image tag entirely (v1.3.50 vs v1.3.67, neither of which
matched what's actually live).
- Reconcile nearle-fiesta.yaml and nearle-jupiter.yaml image tags to
the confirmed-live versions (v1.3.78 / v2.7.55).
- Add allowPrivilegeEscalation:false + drop-all-capabilities to
fiesta/atlantis/jupiter/titan/ariane and the 5 specialized core
workers, which previously ran with no securityContext at all.
- Add terminationGracePeriodSeconds:45 to the worker StatefulSets so
Kubernetes gives the new graceful-shutdown drain (see worker.py
changes) enough time before SIGKILL.
