jakman85andClaude Opus 5
6a8953e288
Minimal prod delivery, authored directly on main (= prod's live revision d1a6628)
rather than by merging work/prod-standup-campaign. That branch carries the DNS apex
inversion, which would have set prod's BASE_DOMAIN to home.keithjackson.dev; prod
runs external-dns with policy: sync and txtOwnerId: prod, so it would have begun
rewriting apex records dev owns while the non-k3s prod still serves 443 there.
Operator sequencing is "dev migrates first". Nothing under clusters/ is touched
here at all, which is what makes this safe to land now.
Also deliberately EXCLUDED, because prod's live cluster-vars has no CNPG_BACKUP_PATH
key and StrictPostBuildSubstitutions would fail the whole tier: the R2 rework of
platform/velero/ and the CNPG destinationPath changes. Those land as
feat/r2-backup-prod together with the endpoint and bucket knobs.
WHAT THIS FIXES
1. Nothing was watching CNPG. All three prod databases have run since bootstrap with
ContinuousArchiving=False and firstRecoverabilityPoint=null, and no alert existed
anywhere. Because archive_mode=on and the archive command fails, Postgres refuses
to recycle WAL: keycloak-db holds every segment ever written (257 MiB on a 5 GiB
volume, ~6.8 MiB/h) — roughly 29 days to a disk-full PANIC on the cluster's SSO
database. Found by hand, 38h late. CNPGWALArchiveBacklog would have fired within
the hour, and evaluates to 3 firing series on prod today.
Rules key on cnpg_collector_* rather than cnpg_pg_stat_archiver_*, because
nextcloud-db's exporter cannot log in to its own database (permission denied for
database "nextcloud", both clusters) so the pg_stat_archiver family does not exist
for it — the obvious metric would have been blind to one database in three.
2. No PVC outside namespace=monitoring had any fill alert. New rules are partitioned
by namespace so they cannot double-fire with MonitoringVolumeFillingUp, pinned to
job="kubelet" (the apiserver copy of the same series reports namespace=default),
and guarded by inodes > 0 — block-mode PVCs such as unifi-rootdisk report
available=0 forever and would otherwise have fired critical permanently.
3. VPNTunnelDown could never clear on a cluster with no gateway node; it has been
firing critical on prod since 2026-09-05. Now guarded so prod self-arms the moment
a node carries the egress-gateway label, with dev's behaviour unchanged today.
4. The Keycloak realm import fails OPEN: an unresolved placeholder is stored
LITERALLY and the import still reports success, then sits at Done=True and never
re-runs. On 2026-09-06 all 12 prod OIDC clients held the literal
$(env:..._CLIENT_SECRET) string and every SSO login failed, while every signal was
green. The CronJob is the level-triggered correction; two alerts make its silence
observable. Repaired live already — this stops it recurring.
Also: the R-NET-04 conformance scan descended into nested agent worktrees under
.claude/, counting 4 egress policies where 1 is authored.
Co-Authored-By: Claude Opus 5 <[email protected]>
115 lines
6.0 KiB
YAML
115 lines
6.0 KiB
YAML
# =============================================================================
|
|
# L2 BEHAVIORAL — split-horizon INTERNAL resolution (R-DNS-01 / R-DNS-03) [RUNNABLE]
|
|
# Catalog: docs/14 "DNS/exposure (R-DNS-*)"
|
|
# Spec: docs/09 "Split-horizon resolution (R-DNS-03)" — internal clients resolve
|
|
# <svc>.${BASE_DOMAIN} to the INTERNAL Gateway VIP, not the public hairpin.
|
|
# Impl: platform/coredns/coredns-custom.yaml (k3s CoreDNS `import *.server`; template plugin
|
|
# answers the whole ${BASE_DOMAIN} zone with ${GATEWAY_VIP}).
|
|
#
|
|
# ENVIRONMENT-AGNOSTIC (R-ENV-02): this test asks the CLUSTER IT IS POINTED AT for its own
|
|
# ${BASE_DOMAIN} and ${GATEWAY_VIP} — the same flux-system/cluster-vars ConfigMap Flux
|
|
# substitutes from — via test-level chainsaw bindings (see `spec.bindings`). Nothing here is
|
|
# per-environment and no flag is passed at run time, so the suite cannot be run against the
|
|
# wrong expectation: dev asserts dev's zone/VIP, prod asserts prod's.
|
|
#
|
|
# GIVEN the split-horizon coredns-custom ConfigMap is reconciled into k3s CoreDNS,
|
|
# WHEN an in-cluster pod resolves a platform hostname over its PUBLIC name
|
|
# (keycloak.${BASE_DOMAIN}) using the default cluster resolver (CoreDNS 10.43.0.10),
|
|
# THEN the answer is EXACTLY that cluster's ${GATEWAY_VIP} (the address coredns-custom is
|
|
# templated with, and the `shared` Gateway's own status address) — NOT a public/off-cluster
|
|
# IP. This is what lets pod->public-hostname traffic hairpin through the
|
|
# Gateway with the correct SNI (*.${BASE_DOMAIN}) so the wildcard cert validates
|
|
# (R-DNS-01 "services reachable at <svc>.${BASE_DOMAIN}"; R-DNS-03 split-horizon).
|
|
#
|
|
# Complements dns-exposure (R-DNS-02, the EXTERNAL/public-record side). Together they cover the
|
|
# split-horizon pair: external answer = public record (R-DNS-02); internal answer = Gateway VIP
|
|
# (R-DNS-01/03).
|
|
#
|
|
# CLEANUP: chainsaw deletes the per-test namespace (Job included) at teardown. Read-only otherwise.
|
|
# chainsaw v0.2.12: Job + status assert; no test-level finally.
|
|
# =============================================================================
|
|
apiVersion: chainsaw.kyverno.io/v1alpha1
|
|
kind: Test
|
|
metadata:
|
|
name: dns-internal-resolution
|
|
labels:
|
|
conformance.home.k3s/R-DNS-01: "true"
|
|
conformance.home.k3s/R-DNS-03: "true"
|
|
annotations:
|
|
conformance.home.k3s/requirements: "R-DNS-01,R-DNS-03"
|
|
conformance.home.k3s/layer: "L2"
|
|
conformance.home.k3s/safety-critical: "false"
|
|
conformance.home.k3s/executable: "true"
|
|
conformance.home.k3s/intent: >-
|
|
An in-cluster pod resolving keycloak.${BASE_DOMAIN} via cluster DNS gets that cluster's
|
|
internal Gateway VIP, not the public hairpin — split-horizon in-cluster resolution.
|
|
spec:
|
|
timeouts:
|
|
apply: 60s
|
|
assert: 2m
|
|
exec: 2m
|
|
# -- R-ENV-02: the environment's identity is READ FROM THE CLUSTER UNDER TEST, never authored
|
|
# here. flux-system/cluster-vars is the ConfigMap Flux itself substitutes from
|
|
# (clusters/base/cluster-vars.yaml + the dev/prod overlay patch), so these bindings are the
|
|
# same source of truth the platform is rendered from. Test-level => in scope for every step
|
|
# (unlike step `outputs`, which are step-scoped).
|
|
bindings:
|
|
- name: baseDomain
|
|
value: (x_k8s_get($client, 'v1', 'ConfigMap', 'flux-system', 'cluster-vars').data.BASE_DOMAIN)
|
|
- name: gatewayVip
|
|
value: (x_k8s_get($client, 'v1', 'ConfigMap', 'flux-system', 'cluster-vars').data.GATEWAY_VIP)
|
|
steps:
|
|
- name: then-cluster-dns-answers-with-the-internal-gateway-vip
|
|
try:
|
|
- apply:
|
|
resource:
|
|
apiVersion: batch/v1
|
|
kind: Job
|
|
metadata:
|
|
name: dns-internal-probe
|
|
spec:
|
|
backoffLimit: 2
|
|
activeDeadlineSeconds: 100
|
|
template:
|
|
spec:
|
|
restartPolicy: Never
|
|
containers:
|
|
- name: probe
|
|
image: busybox:1.36.1
|
|
# The cluster's OWN identity, resolved from flux-system/cluster-vars by the
|
|
# test-level bindings — never a literal (R-ENV-02).
|
|
env:
|
|
- name: BASE_DOMAIN
|
|
value: ($baseDomain)
|
|
- name: GATEWAY_VIP
|
|
value: ($gatewayVip)
|
|
command:
|
|
- sh
|
|
- -c
|
|
- |
|
|
set -u
|
|
host="keycloak.${BASE_DOMAIN}"
|
|
# Resolve via the DEFAULT cluster resolver (CoreDNS, /etc/resolv.conf ->
|
|
# 10.43.0.10). Split-horizon must answer with THIS cluster's Gateway VIP
|
|
# (the address coredns-custom is templated with), proving pods hairpin
|
|
# through the Gateway. Exact match on ${GATEWAY_VIP}: stricter than the
|
|
# old hardcoded 192.168.1.53-.60 range AND correct in both environments.
|
|
out="$(nslookup "$host" 2>/dev/null)"
|
|
echo "$out"
|
|
ip="$(echo "$out" | awk '/^Address: /{print $2}' | grep -Fx "${GATEWAY_VIP}" | head -1)"
|
|
if [ -n "$ip" ]; then
|
|
echo "INTERNAL_RESOLUTION_OK $host -> $ip (in-cluster Gateway VIP)"
|
|
exit 0
|
|
fi
|
|
echo "FAIL: $host did not resolve to this cluster's Gateway VIP (${GATEWAY_VIP});"
|
|
echo " split-horizon coredns-custom not in effect (got a public/off-cluster answer)."
|
|
exit 1
|
|
- assert:
|
|
resource:
|
|
apiVersion: batch/v1
|
|
kind: Job
|
|
metadata:
|
|
name: dns-internal-probe
|
|
status:
|
|
succeeded: 1
|