jakman85andClaude Opus 5
a92fa11701
An alert rule whose metric is never scraped is not a broken alert, it is an ABSENT
alert that reads as a present one. `longhorn_volume_robustness == 2` over a TSDB
with no longhorn_* series returns no vector: permanently quiet, permanently
"healthy", indistinguishable from a correct rule. It survives review (the PromQL is
right), it survives `promtool test rules` (the unit test supplies the series
itself), and it appears on no dashboard.
VERIFIED ABSENT ON PROD before this change (re-verified 2026-09-13 21:0xZ, ~15h
after the first audit, because a stale reading is not evidence):
count(longhorn_volume_robustness) => (empty) 4 rules inert
count({__name__=~"certmanager.*"}) => (empty) 3 rules inert
count(hubble_drop_total) / {__name__=~"cilium.*"} => (empty) 1 rule inert
count by (job) (up) => 20 jobs, none of them
Longhorn, cert-manager,
Cilium or Hubble
Consequence: storage health was unmonitored on the cluster whose prod-1 had just
had a failing SSD, and TLS expiry was unmonitored with the live `wildcard-home`
wildcard 89 days from expiry on an ACME DNS-01 renewal that can fail silently.
WHAT THIS CHANGES
* platform/longhorn: `metrics.serviceMonitor.enabled: true` (the supported chart
way — rendering longhorn 1.12.1 emits exactly 1 ServiceMonitor with the flag and
0 without), plus networkpolicy-metrics.yaml. The NetworkPolicy is LOAD-BEARING,
not belt-and-braces: measured with the monitor applied and no allowance, all 3
targets were DISCOVERED and every scrape failed `context deadline exceeded`
after 10.0s — a drop, not a refusal — while curling the endpoint from inside
the manager pod returned 200 in 0.031s. Longhorn 1.12.1 ships 6 NetworkPolicies
of its own, guarded on `networkPolicies.restrictInternalTraffic` (default TRUE)
and NOT on `networkPolicies.enabled`, so they are effectively unconditional and
the manager policy admits only same-namespace Longhorn components. The one-line
alternative (restrictInternalTraffic: false) was rejected: it deletes the
internal-traffic restriction across all six components to open one port to one
namespace, trading an R-SEC-02 control for an observability one when the two do
not conflict. This policy is Ingress-only, ns monitoring -> app=longhorn-manager,
TCP 9500 only; 9502/9503 (webhook, recovery backend) stay closed.
* platform/cert-manager: `prometheus.servicemonitor.enabled: true`. Chart default
already serves 9402; only the monitor was missing. No NetworkPolicy needed —
neither ns cert-manager nor ns monitoring carries one. NOT paired with
prometheus.podmonitor.enabled: the chart calls `fail` if both are set.
* platform/kubevirt: KubeVirtVMLaunchFailing queried
`kubevirt_vmi_phase_count{phase="Failed"}`, which DOES NOT EXIST on the deployed
KubeVirt — a different flavour of the same defect, with the scrape target up and
healthy. Prod has only `kubevirt_vmi_info` and `kubevirt_vmi_number_of_outdated`;
the only phase-bearing names are the phase_transition_time histograms. Corrected
to `kubevirt_vmi_info{phase="failed"}` — LOWERCASE, because KubeVirt lowercases
the value (observed `phase="running"` for the unifi VMI) and a capital-F matcher
would select nothing and be exactly as inert.
PROVEN, NOT ASSERTED. The rendered ServiceMonitors were applied to prod by hand and
the metrics observed to land, then REMOVED again so the merge's Helm upgrade creates
and owns them rather than colliding with an unowned object:
count(up{job="longhorn-backend"}) => 3 (all three managers, up==1)
count({__name__=~"longhorn.*"}) => 8717 (was empty)
count(longhorn_volume_robustness) => 44
up{namespace="cert-manager"} => 3 targets, all 1
count({__name__=~"certmanager.*"}) => 104 (was empty)
wildcard-home days to expiry => 89.2, ready_status{condition="True"}=1
All 58 alert rules in the repo were then evaluated against prod one by one: every
one returns a result without error, and exactly ONE fires.
TRIAGE OF THE ONE NEWLY-FIRING ALERT — A TRUE POSITIVE, NOT TUNED AWAY.
LonghornVolumeCapacityHigh on the Prometheus TSDB volume:
`longhorn_volume_actual_size_bytes / longhorn_volume_capacity_bytes` = 0.971
(48.6 GiB allocated of 50 GiB) while the kubelet's view of the same filesystem is
only 20.4% used. Longhorn is not reclaiming freed blocks, so ~48.6 GiB x 3 replicas
is committed for ~10 GiB of real data. On a cluster whose failure mode is disk
pressure and whose etcd WAL shares a spindle with Longhorn, that is worth knowing
and was invisible until now. The threshold is left alone; the follow-up is
filesystem trim, not a bigger number.
THE GUARD (conformance/tests/l0-static/alert-metrics-have-scrape-targets)
Whole-repo L0 invariant, 118 assertions. It crosses from each alert to the scrape
config that makes its metric exist: every metric name is claimed by exactly one
declared owner, and each owner's enabling declaration is re-proven against the
working tree on every gate run, including the negative conditions that defeat an
enablement without deleting it. `up{job="X"}` is audited on its own axis, because
`up` always exists: where this repo authors a Service and a ServiceMonitor in the
same file, the monitor's endpoint port must be a port that Service defines.
It DECLARES conformance.home.k3s/scope: "whole-repo" and adds NO ci/l0.sh stage —
the whole-repo-invariants-wired meta-test fails the gate if one reappears.
PROVEN ABLE TO FAIL, five ways, each reverted:
1. an alert on a fabricated metric -> FAIL "claimed by NO entry"
2. longhorn metrics.serviceMonitor removed -> FAIL "no longer matches"
3. networkpolicy-metrics.yaml unwired -> FAIL "the monitor alone is NOT enough"
4. cert-manager podmonitor also enabled -> FAIL "FORBIDDEN pattern"
5. a register entry made stale / removed -> FAIL in both directions
TWO SHRINK-ONLY REGISTERS, because a blanket fail would make the gate red until
work that is not this change lands, and a disabled gate protects nothing:
* KNOWN_UNWIRED_METRICS (1): hubble_drop_total. See below.
* KNOWN_UNWIRED_JOBS (12): every apps/* `<app>-down` SLI rule is ALSO inert, and
this was found by the re-audit, not previously known. The rendered
ServiceMonitors name `port: metrics` and NOT ONE of the twelve app Services
defines a port of that name (zero occurrences of `name: metrics` across all
twelve manifests); prod confirms none of the twelve jobs appears in `up`. This
is the nastiest variant of the class: `up` exists in abundance so nothing looks
wrong, and PrometheusTargetsMissing cannot see it either, because a target that
was never discovered has no up series to be 0. Closing it needs an exporter per
app (Plex, Pi-hole, Jackett, qBittorrent, Sonarr, Radarr, Seerr speak no
Prometheus) — except keycloak, where the live jobs are keycloak-service /
keycloak-discovery and the rule is a rename away from being live.
NOT DONE HERE, DELIBERATELY: Hubble. hubble.metrics.enabled RECONFIGURES THE CNI —
it opens 9965 on every cilium-agent and rolls the DaemonSet across all three nodes,
serialised by this repo's own maxUnavailable: 1 (set after the 2026-09-01
k8sServiceHost outage, where a CNI roll deadlocked 2 of 3 agents). It also needs a
matching edit to roles/cilium/templates/cilium-values.yaml.j2 in the same commit,
because platform/cilium/helmrelease.yaml carries a [parity] contract against it and
a rebuilt node would otherwise bootstrap a CNI config that disagrees with the
reconciled one. Recommended as its own change on the cilium Kustomization, with the
Ansible parity edit, a narrow labelsContext, and a one-node canary — not smuggled in
here.
ALSO FOUND, NOT FIXED HERE: kube-state-metrics is OOMKilled every ~10 minutes on
prod (90 restarts in 16h against a 128Mi limit) while its logs show the
custom-resource-state factory reloading every 3 seconds. Thirteen alert rules read
kube_* metrics, so those rules are INTERMITTENTLY inert — the same defect class,
arriving through a crashloop rather than a missing monitor. Needs its own change.
Co-Authored-By: Claude Opus 5 <[email protected]>
159 lines
10 KiB
YAML
159 lines
10 KiB
YAML
---
|
|
# Longhorn — distributed replicated block storage across the 3 nodes (ADR-0003).
|
|
# Serves R-STATE-01 (state survives destroy/rebuild) and R-STATE-02 (3-way replication).
|
|
#
|
|
# Dev-appropriate values. Environment differences enter ONLY via $${...} substitution from
|
|
# cluster-vars (R-ENV-02); everything else is identical base config for dev and prod.
|
|
apiVersion: helm.toolkit.fluxcd.io/v2
|
|
kind: HelmRelease
|
|
metadata:
|
|
name: longhorn
|
|
namespace: longhorn-system
|
|
spec:
|
|
interval: 30m
|
|
# Longhorn upgrades roll every workload + engine image on slow HDD-backed nodes and can exceed the
|
|
# 5m Helm default; a short timeout caused a W5 mid-upgrade wedge (data migrated to the new minor but
|
|
# workloads rolled back, and the new-version data refuses the old-version manager). Give it room.
|
|
timeout: 15m0s
|
|
chart:
|
|
spec:
|
|
chart: longhorn
|
|
version: ">=1.12.0 <1.13.0" # W5 upgrade 1.11->1.12 (resolves 1.12.1; supports k8s 1.36). Renovate maintains
|
|
sourceRef:
|
|
kind: HelmRepository
|
|
name: longhorn
|
|
namespace: longhorn-system
|
|
install:
|
|
createNamespace: false # namespace.yaml owns it (GitOps-managed, labelled)
|
|
remediation:
|
|
retries: 3
|
|
upgrade:
|
|
remediation:
|
|
retries: 3
|
|
# NEVER auto-rollback a Longhorn minor upgrade: once the on-disk data migrates to the new minor,
|
|
# the older manager fatally refuses it ("downgrading from vX to vY is not supported"). Rolling
|
|
# back the workloads while data is already migrated wedges the cluster. Fail forward instead.
|
|
remediateLastFailure: false
|
|
values:
|
|
# We own the StorageClasses (storageclasses.yaml) so the default 3-replica class and the bulk
|
|
# tier are explicit and GitOps-managed. Disable the chart's built-in default SC so there is
|
|
# exactly one default-class annotation in the cluster.
|
|
persistence:
|
|
defaultClass: false
|
|
|
|
defaultSettings:
|
|
# --- data path (R-STATE-02, R-ENV-02) ---
|
|
# The on-node directory every replica is written into. A cluster-vars KNOB, not a literal,
|
|
# because the two environments mount their data disk differently and a hardcoded path would
|
|
# put one of them on the OS filesystem:
|
|
# dev /var/lib/longhorn — a dedicated ext4 disk (/dev/vdb) mounted there by
|
|
# roles/node_baseline. The mount root IS the dataPath: one disk,
|
|
# one tenant. 18 live volumes; this value must never change.
|
|
# prod /data/longhorn — a SUBDIRECTORY of the claimed ~424 GB partition mounted at
|
|
# /data, whose sibling /data/k3s holds k3s's entire data dir
|
|
# (etcd + the containerd image store, moved off the 11.8 GiB
|
|
# /var). Two tenants, so Longhorn gets a subdirectory rather
|
|
# than the mount root. infra/PROD-PROVISIONING.md §2.
|
|
# [parity] MUST equal roles/node_baseline's node_baseline_longhorn_data_path for the same
|
|
# environment — Ansible creates and mounts the directory, Helm tells Longhorn to use it.
|
|
# Set in BOTH overlays (clusters/dev, clusters/prod); base default is the dev value.
|
|
defaultDataPath: "${LONGHORN_DATA_PATH}"
|
|
|
|
# --- replication (R-STATE-02) ---
|
|
# Global fallback replica count. The `longhorn` StorageClass sets numberOfReplicas explicitly
|
|
# (authoritative); this keeps any volume created without an explicit count consistent.
|
|
# Substituted from cluster-vars.longhorn_default_replica_count (=3 in dev).
|
|
defaultReplicaCount: ${LONGHORN_DEFAULT_REPLICA_COUNT}
|
|
|
|
# --- replica anti-affinity across nodes (R-STATE-02 durability guarantee) ---
|
|
# HARD anti-affinity: never place two replicas of the same volume on the same node, so a
|
|
# single node loss cannot take down more than one replica. On a 3-node / 3-replica cluster
|
|
# this yields exactly one replica per node.
|
|
replicaSoftAntiAffinity: false
|
|
# Single zone (3 mini-PCs, no failure zones): allow replicas within the one zone.
|
|
replicaZoneSoftAntiAffinity: true
|
|
# When a downed node returns, rebalance replicas back toward one-per-node (best-effort so it
|
|
# never blocks scheduling).
|
|
replicaAutoBalance: "best-effort"
|
|
|
|
# --- rebuild I/O throttle for slow disks (etcd WAL fsync protection) ---
|
|
# The k3s VM disks sit on HDDs, so a full-replica rebuild's sequential I/O can saturate the
|
|
# spindle and starve etcd's WAL fsync — which crash-loops etcd and flaps the kube-vip VIP.
|
|
# Cap Longhorn to rebuilding ONE replica per node at a time so a node bounce / replica
|
|
# replenishment can never storm the HDDs. Deliberately conservative for this slow-disk cluster
|
|
# (Longhorn's chart default is higher). Pair with replica-replenishment-wait-interval=600
|
|
# (Longhorn default) so brief node bounces don't trigger replenishment at all.
|
|
concurrentReplicaRebuildPerNodeLimit: 1
|
|
|
|
# --- capacity guardrails (docs/04 "reserve disk headroom") ---
|
|
# Reserve 20% of the dedicated data disk so a node never fills to 100% (which wedges the
|
|
# kubelet and Longhorn alike). Dev disks are small, so headroom matters.
|
|
storageReservedPercentageForDefaultDisk: "20"
|
|
# Stop scheduling new replicas onto a disk once free space drops below 25%.
|
|
storageMinimalAvailablePercentage: "25"
|
|
# Dev: 200% thin over-provisioning — config-only volumes use little actual space, so the small
|
|
# 39Gi data disks host the full app catalog. Physical fill risk is low (watch monitoring PVCs).
|
|
# Prod-grade alternative: grow the Longhorn data disks and return to 1:1 (WS-STORAGE).
|
|
storageOverProvisioningPercentage: "200"
|
|
|
|
# --- misc dev-appropriate defaults ---
|
|
# Keep replicas spread for durability rather than co-locating with the pod.
|
|
defaultDataLocality: "disabled"
|
|
# Reclaim replica space of deleted/failed volumes promptly in dev.
|
|
orphanAutoDeletion: true
|
|
|
|
# --- SCRAPE THE STORAGE LAYER, or every Longhorn alert is INERT (added 2026-09-13) ----------
|
|
# platform/monitoring/alerts.yaml ships LonghornVolumeDegraded (critical),
|
|
# LonghornVolumeFaulted (critical), LonghornVolumeCapacityHigh and LonghornNodeStorageHigh —
|
|
# and NONE of them could ever fire, because the chart defaults
|
|
# `metrics.serviceMonitor.enabled: false` and no ServiceMonitor existed in ns longhorn-system.
|
|
# VERIFIED live on prod 2026-09-13: `count by (job) (up)` listed 19 jobs and none was Longhorn;
|
|
# `count(longhorn_volume_robustness)` and `count({__name__=~"longhorn.*"})` both returned NOTHING.
|
|
# So the entire longhorn_* family was absent from the TSDB and four alert rules were decoration.
|
|
#
|
|
# WHY THIS IS PART OF THE STORAGE PATH, NOT AN OBSERVABILITY NICE-TO-HAVE: this cluster's
|
|
# recurring failure mode is a disk. prod-1's failing SSD starved etcd's WAL fsync and wedged the
|
|
# node (DECISIONS-LOG; the prod-1 drive replacement), and a degraded/faulted replica is precisely
|
|
# the first symptom Longhorn publishes. R-STATE-02 promises 3-way replication survives a node
|
|
# loss; with no scrape, a volume can sit at 2 replicas — or faulted — indefinitely, unalerted.
|
|
# platform/monitoring/pvc-capacity-alerts.yaml's header already flagged this exact gap as "a
|
|
# separate defect (a missing Longhorn ServiceMonitor/metrics config)". This closes it.
|
|
#
|
|
# Chart mechanics, verified by rendering longhorn 1.12.1 (the version this range resolves to):
|
|
# * templates/servicemonitor.yaml is gated on `{{- if .Values.metrics.serviceMonitor.enabled }}`
|
|
# and NOTHING else — no `.Capabilities` probe. Measured: `helm template --set
|
|
# metrics.serviceMonitor.enabled=true` emits exactly 1 ServiceMonitor, and 0 without the flag.
|
|
# So this line is both necessary and sufficient; unlike velero's, it needs no API capability.
|
|
# * the rendered ServiceMonitor selects `app: longhorn-manager` in the release namespace on the
|
|
# port named `manager`. CONFIRMED against prod: Service/longhorn-backend carries
|
|
# `app: longhorn-manager` and a port `manager` (9500) — the selector resolves to real
|
|
# endpoints, it is not a dangling monitor (which is how the apps/ SLI monitors fail: they name
|
|
# a `metrics` port their Service does not have, so they discover zero targets and their
|
|
# `up{job=...}` alerts are inert in the very same way).
|
|
# * no `additionalLabels` needed: platform/monitoring sets serviceMonitorSelectorNilUsesHelmValues
|
|
# false, so Prometheus selects every ServiceMonitor regardless of labels.
|
|
# Cost: one scrape target per manager pod (3), the longhorn_* families — low hundreds of series
|
|
# on this cluster's volume count.
|
|
#
|
|
# PROVEN END TO END ON PROD, 2026-09-13 21:0xZ. The rendered ServiceMonitor was applied to prod
|
|
# by hand (then removed again, so the merge's Helm upgrade creates and OWNS it rather than
|
|
# colliding with an unowned object) with this file's networkpolicy-metrics.yaml already in place:
|
|
# count(up{job="longhorn-backend"}) => 3 (all three managers, up==1)
|
|
# count({__name__=~"longhorn.*"}) => 8717 (was EMPTY)
|
|
# count(longhorn_volume_robustness) => 44
|
|
# count_values("v", longhorn_volume_robustness) => 24 healthy(1), 72 unknown(0, detached)
|
|
# and the four alert rules went from "no series, cannot fire" to evaluating for real — one of
|
|
# them immediately found a true positive (LonghornVolumeCapacityHigh on the Prometheus TSDB
|
|
# volume at 97.1% ALLOCATED while its filesystem is only 20.4% used: Longhorn is not reclaiming
|
|
# freed blocks. Reported, NOT silenced by moving the threshold).
|
|
metrics:
|
|
serviceMonitor:
|
|
enabled: true
|
|
|
|
# Longhorn's own manager/UI stay in-cluster; UI exposure (if any) is gated behind
|
|
# oauth2-proxy forward-auth (ADR-0014) by WS-SEC — not exposed here.
|
|
|
|
# TODO(WS-STORAGE, Phase 4): off-cluster backup target (defaultSettings.backupTarget +
|
|
# backupTargetCredentialSecret) is configured alongside Velero (ADR-0003, R-STATE-03).
|
|
# Credentials referenced BY NAME only, SOPS-encrypted (WS-SEC) — never inline here.
|