From b4493b3d0fd09c7ee65ce78007d8b89caf19591b Mon Sep 17 00:00:00 2001 From: dingo Date: Wed, 30 Sep 2026 06:16:36 -0400 Subject: [PATCH] Add M8.5 observability: pinned Prometheus, tested alerts, capacity metrics An AEvolution-only Prometheus 3.13.3 and Alertmanager 0.34.1, pinned by digest and delivered by Argo (ADR-0072), scrape each region's operations Service through static targets (no API access, no RBAC); the region namespaces admit port 9100 from Prometheus only. Sixteen alert rules cover authority, lease, commits, backlog and object capacity and are proven by promtool injection tests in CI with tools copied from the same pinned images. Two new gauges expose object volume free space and the reserve. Alerts route to a null receiver until a destination is chosen; no broker is adopted and the database outbox stays. Co-Authored-By: Claude Opus 5.5 --- apps/evo-server/main.cpp | 2 + deploy/argocd/monitoring/application.yaml | 19 ++ deploy/argocd/monitoring/project.yaml | 23 ++ deploy/argocd/platform/project.yaml | 2 + deploy/docker/Dockerfile | 5 + .../kubernetes/monitoring/alertmanager.yaml | 90 +++++++ .../monitoring/config/alertmanager.yml | 17 ++ .../monitoring/config/evo-region.rules.yml | 121 +++++++++ .../monitoring/config/prometheus.yml | 33 +++ .../kubernetes/monitoring/kustomization.yaml | 20 ++ deploy/kubernetes/monitoring/prometheus.yaml | 115 +++++++++ .../tests/evo-region.rules.test.yml | 240 ++++++++++++++++++ .../platform/cluster/kustomization.yaml | 1 + .../platform/monitoring/guardrails.yaml | 65 +++++ .../platform/monitoring/kustomization.yaml | 13 + .../platform/monitoring/namespace.yaml | 10 + .../kubernetes/region/base/kustomization.yaml | 1 + .../kubernetes/region/base/scrape-policy.yaml | 21 ++ docs/BACKLOG.md | 2 + docs/ROADMAP.md | 8 + docs/STATUS.md | 19 ++ docs/adr/0066-operations-endpoints.md | 7 + docs/adr/0072-observability.md | 125 +++++++++ docs/adr/README.md | 1 + docs/engineering/VALIDATION.md | 33 +++ docs/infrastructure/ENVIRONMENTS.md | 21 ++ include/evo/server/ops.hpp | 3 + include/evo/server/region_host.hpp | 4 + src/server/ops.cpp | 4 + tests/ops_tests.cpp | 6 +- tests/region_ops_tests.cpp | 3 + tools/ci/deploy-policy-check.py | 27 +- tools/ci/monitoring-check.sh | 15 ++ 33 files changed, 1068 insertions(+), 8 deletions(-) create mode 100644 deploy/argocd/monitoring/application.yaml create mode 100644 deploy/argocd/monitoring/project.yaml create mode 100644 deploy/kubernetes/monitoring/alertmanager.yaml create mode 100644 deploy/kubernetes/monitoring/config/alertmanager.yml create mode 100644 deploy/kubernetes/monitoring/config/evo-region.rules.yml create mode 100644 deploy/kubernetes/monitoring/config/prometheus.yml create mode 100644 deploy/kubernetes/monitoring/kustomization.yaml create mode 100644 deploy/kubernetes/monitoring/prometheus.yaml create mode 100644 deploy/kubernetes/monitoring/tests/evo-region.rules.test.yml create mode 100644 deploy/kubernetes/platform/monitoring/guardrails.yaml create mode 100644 deploy/kubernetes/platform/monitoring/kustomization.yaml create mode 100644 deploy/kubernetes/platform/monitoring/namespace.yaml create mode 100644 deploy/kubernetes/region/base/scrape-policy.yaml create mode 100644 docs/adr/0072-observability.md create mode 100644 tools/ci/monitoring-check.sh diff --git a/apps/evo-server/main.cpp b/apps/evo-server/main.cpp index cff6a50..2c5a261 100644 --- a/apps/evo-server/main.cpp +++ b/apps/evo-server/main.cpp @@ -310,6 +310,7 @@ template int serve(Options& o) { server::OpsCounters counters; counters.queue_capacity_inputs = o.host.max_queued_inputs; counters.queue_capacity_bytes = o.host.max_queued_bytes; + counters.objects_reserve_bytes = o.objects_reserve_mib * 1024 * 1024; std::uint64_t last_sequence = 0; auto last_checkpoint = std::chrono::steady_clock::now(); bool recovered = false; @@ -326,6 +327,7 @@ template int serve(Options& o) { static_cast(std::chrono::duration_cast( std::chrono::steady_clock::now() - last_checkpoint) .count()); + counters.objects_available_bytes = host.objects().available_bytes().value_or(0); return {200, server::render_metrics(s, counters), "text/plain; version=0.0.4; charset=utf-8"}; } diff --git a/deploy/argocd/monitoring/application.yaml b/deploy/argocd/monitoring/application.yaml new file mode 100644 index 0000000..c42c084 --- /dev/null +++ b/deploy/argocd/monitoring/application.yaml @@ -0,0 +1,19 @@ +# M8.5 monitoring application (ADR-0072). Manual sync like the other AEvolution +# workloads; it holds no world data. +apiVersion: argoproj.io/v1alpha1 +kind: Application +metadata: + name: aevolution-monitoring + namespace: argocd +spec: + project: aevolution-monitoring + source: + repoURL: https://git.aeternalisgames.com/dingo/AEvolution.git + targetRevision: main + path: deploy/kubernetes/monitoring + destination: + server: https://kubernetes.default.svc + namespace: aevolution-monitoring + syncPolicy: + syncOptions: + - FailOnSharedResource=true diff --git a/deploy/argocd/monitoring/project.yaml b/deploy/argocd/monitoring/project.yaml new file mode 100644 index 0000000..396c079 --- /dev/null +++ b/deploy/argocd/monitoring/project.yaml @@ -0,0 +1,23 @@ +# M8.5 monitoring project (ADR-0072). Apply by hand once: +# kubectl apply -f deploy/argocd/monitoring/ +apiVersion: argoproj.io/v1alpha1 +kind: AppProject +metadata: + name: aevolution-monitoring + namespace: argocd +spec: + description: AEvolution metrics and alerting (aevolution-monitoring only) + sourceRepos: + - https://git.aeternalisgames.com/dingo/AEvolution.git + destinations: + - server: https://kubernetes.default.svc + namespace: aevolution-monitoring + clusterResourceWhitelist: [] + namespaceResourceWhitelist: + - {group: "", kind: Service} + - {group: "", kind: ConfigMap} + - {group: "", kind: PersistentVolumeClaim} + - {group: apps, kind: Deployment} + - {group: networking.k8s.io, kind: NetworkPolicy} + orphanedResources: + warn: true diff --git a/deploy/argocd/platform/project.yaml b/deploy/argocd/platform/project.yaml index ea23ca6..911be85 100644 --- a/deploy/argocd/platform/project.yaml +++ b/deploy/argocd/platform/project.yaml @@ -18,6 +18,8 @@ spec: namespace: evolution-stage - server: https://kubernetes.default.svc namespace: evolution-world0 + - server: https://kubernetes.default.svc + namespace: aevolution-monitoring clusterResourceWhitelist: - {group: "", kind: Namespace} - {group: storage.k8s.io, kind: StorageClass} diff --git a/deploy/docker/Dockerfile b/deploy/docker/Dockerfile index efbf777..7ad688c 100644 --- a/deploy/docker/Dockerfile +++ b/deploy/docker/Dockerfile @@ -57,6 +57,11 @@ COPY deploy/ deploy/ COPY tools/ci/deploy-policy-check.py tools/ci/ COPY tools/ops/ tools/ops/ RUN python3 tools/ci/deploy-policy-check.py +# M8.5: monitoring configuration and alert tests with the pinned images' own tools. +COPY --from=quay.io/prometheus/prometheus@sha256:6976aa8a60fec930796ce5772b8d12da7a318a5daa8d40d69c5c7819a05eeed7 /bin/promtool /usr/local/bin/promtool +COPY --from=quay.io/prometheus/alertmanager@sha256:e9733bafb1bdef9b00e25a21f8f99dc26a22224bf16641ad754d1649f4c3357a /bin/amtool /usr/local/bin/amtool +COPY tools/ci/monitoring-check.sh tools/ci/ +RUN bash tools/ci/monitoring-check.sh FROM ci-toolchain AS postgres-build RUN apt-get update && apt-get install -y --no-install-recommends \ diff --git a/deploy/kubernetes/monitoring/alertmanager.yaml b/deploy/kubernetes/monitoring/alertmanager.yaml new file mode 100644 index 0000000..03892e8 --- /dev/null +++ b/deploy/kubernetes/monitoring/alertmanager.yaml @@ -0,0 +1,90 @@ +# M8.5 Alertmanager (ADR-0072), pinned by digest. Every alert routes to the null +# receiver until the maintainer chooses a destination (docs/BACKLOG.md); alerts are +# still visible in its API and UI. Silences live in memory (emptyDir). +apiVersion: apps/v1 +kind: Deployment +metadata: + name: alertmanager + labels: + app.kubernetes.io/name: alertmanager +spec: + replicas: 1 + strategy: + type: Recreate + selector: + matchLabels: + app.kubernetes.io/name: alertmanager + template: + metadata: + labels: + app.kubernetes.io/name: alertmanager + app.kubernetes.io/part-of: aevolution + spec: + serviceAccountName: aevolution-alertmanager + automountServiceAccountToken: false + securityContext: + runAsNonRoot: true + runAsUser: 65534 + runAsGroup: 65534 + fsGroup: 65534 + seccompProfile: {type: RuntimeDefault} + containers: + - name: alertmanager + image: quay.io/prometheus/alertmanager:v0.34.1@sha256:e9733bafb1bdef9b00e25a21f8f99dc26a22224bf16641ad754d1649f4c3357a + args: + - --config.file=/etc/alertmanager/alertmanager.yml + - --storage.path=/alertmanager + - --web.listen-address=:9093 + - --cluster.listen-address= # Single instance: no gossip port. + ports: + - {name: web, containerPort: 9093} + readinessProbe: + httpGet: {path: /-/ready, port: web} + periodSeconds: 10 + livenessProbe: + httpGet: {path: /-/healthy, port: web} + periodSeconds: 20 + resources: + requests: {cpu: 50m, memory: 64Mi} + limits: {memory: 128Mi} + securityContext: + allowPrivilegeEscalation: false + readOnlyRootFilesystem: true + capabilities: {drop: [ALL]} + volumeMounts: + - {name: config, mountPath: /etc/alertmanager, readOnly: true} + - {name: data, mountPath: /alertmanager} + volumes: + - name: config + configMap: + name: alertmanager-config + - name: data + emptyDir: {sizeLimit: 64Mi} +--- +apiVersion: v1 +kind: Service +metadata: + name: alertmanager +spec: + type: ClusterIP + selector: + app.kubernetes.io/name: alertmanager + ports: + - {name: web, port: 9093, targetPort: web} +--- +apiVersion: networking.k8s.io/v1 +kind: NetworkPolicy +metadata: + name: alertmanager +spec: + podSelector: + matchLabels: + app.kubernetes.io/name: alertmanager + policyTypes: [Ingress] + ingress: + - from: + - podSelector: + matchLabels: + app.kubernetes.io/name: prometheus + ports: + - {protocol: TCP, port: 9093} diff --git a/deploy/kubernetes/monitoring/config/alertmanager.yml b/deploy/kubernetes/monitoring/config/alertmanager.yml new file mode 100644 index 0000000..26fcf47 --- /dev/null +++ b/deploy/kubernetes/monitoring/config/alertmanager.yml @@ -0,0 +1,17 @@ +# M8.5 Alertmanager routing (ADR-0072). No destination is chosen yet +# (docs/BACKLOG.md): every alert routes to the null receiver, which keeps it visible +# in Alertmanager's API and UI without delivering it. Adding a receiver later is a +# reviewed change here plus a Secret for its credentials, never a credential in Git. +route: + receiver: "null" + group_by: [alertname, environment] + group_wait: 30s + group_interval: 5m + repeat_interval: 4h +receivers: + - name: "null" +inhibit_rules: + # A region that is down cannot also meaningfully be not ready or stale. + - source_matchers: [alertname = EvoRegionDown] + target_matchers: [alertname =~ "EvoRegion(NotReady|LeaseMargin|CheckpointStale)"] + equal: [environment] diff --git a/deploy/kubernetes/monitoring/config/evo-region.rules.yml b/deploy/kubernetes/monitoring/config/evo-region.rules.yml new file mode 100644 index 0000000..95a0ac5 --- /dev/null +++ b/deploy/kubernetes/monitoring/config/evo-region.rules.yml @@ -0,0 +1,121 @@ +# M8.5 region host alerts (ADR-0072). Thresholds follow the measured behaviour: +# the production lease is 20 s renewed every 5 s (15-20 s remaining when healthy); +# batches are at most 64 ticks with a 1 s flush; an idle region (no inputs, no +# gateway yet) always accrues tick debt, so debt and staleness only alert while +# inputs are waiting. Unit tests: evo-region.rules.test.yml (promtool). +groups: + - name: evo-region-authority + rules: + - alert: EvoRegionDown + expr: up{job="evo-region"} == 0 + for: 2m + labels: {severity: critical} + annotations: + summary: "Region {{ $labels.environment }}/{{ $labels.world }}/{{ $labels.region }} is not scrapeable" + description: "No operations endpoint answers. The process or its pod is gone; a forced deletion recovers in about 20 s, so 2 minutes is a real outage." + - alert: EvoRegionNotReady + expr: evo_region_ready == 0 + for: 2m + labels: {severity: critical} + annotations: + summary: "Region {{ $labels.environment }} has no authoritative host" + description: "The process is alive but holds no lease or has not verified its head. Check the phase metric and the host log." + - alert: EvoRegionQuarantined + expr: evo_region_phase{phase="quarantined"} == 1 + for: 1m + labels: {severity: critical} + annotations: + summary: "Region {{ $labels.environment }} is quarantined" + description: "Committed state or history failed verification. The region stays unavailable; never re-provision it. Audit with evo-world-clone audit." + - alert: EvoRegionLeaseMargin + expr: evo_region_ready == 1 and evo_region_lease_remaining_milliseconds < 10000 + for: 1m + labels: {severity: warning} + annotations: + summary: "Region {{ $labels.environment }} lease is not being renewed" + description: "Healthy renewals keep 15-20 s of lease. Under 10 s for a minute means renewals are failing; the host will fence itself." + - alert: EvoRegionFenced + expr: increase(evo_region_fences_total[10m]) > 0 + labels: {severity: warning} + annotations: + summary: "Region {{ $labels.environment }} host lost authority" + description: "A host was fenced within the last 10 minutes. Expected once after a forced deletion's predecessor resumes; repeated fences point to database or clock trouble." + - name: evo-region-progress + rules: + - alert: EvoRegionStoreFailures + expr: increase(evo_region_store_failures_total[5m]) > 3 + labels: {severity: warning} + annotations: + summary: "Region {{ $labels.environment }} storage or database failures" + description: "Publication or database calls are failing; progress pauses and nothing is acknowledged until they succeed." + - alert: EvoRegionUnknownOutcomes + expr: increase(evo_region_unknown_outcomes_total[15m]) > 0 + labels: {severity: warning} + annotations: + summary: "Region {{ $labels.environment }} had commits with unknown outcomes" + description: "Commits timed out or lost their connection and were resolved by exact bytes. Occasional is safe; frequent indicates database latency near the 5 s commit deadline." + - alert: EvoRegionCommitLag + expr: evo_region_commit_lag_ticks > 256 + for: 2m + labels: {severity: warning} + annotations: + summary: "Region {{ $labels.environment }} is not committing what it simulates" + description: "More than four full batches are simulated but not durable; acknowledgements are waiting." + - alert: EvoRegionTickDebt + expr: evo_region_tick_debt > 8 and evo_region_queued_inputs > 0 + for: 2m + labels: {severity: warning} + annotations: + summary: "Region {{ $labels.environment }} is behind real time with inputs waiting" + description: "Due ticks are not being simulated although inputs are queued (catch-up is 8 ticks per step)." + - alert: EvoRegionCheckpointStale + expr: evo_region_checkpoint_age_milliseconds > 300000 and (evo_region_queued_inputs > 0 or evo_region_commit_lag_ticks > 0) + for: 1m + labels: {severity: warning} + annotations: + summary: "Region {{ $labels.environment }} has not committed for 5 minutes with work pending" + description: "An idle region does not commit; this fires only while inputs or simulated ticks wait for durability." + - alert: EvoRegionQueueSaturated + expr: evo_region_queued_inputs >= 0.9 * evo_region_queue_capacity_inputs + for: 2m + labels: {severity: warning} + annotations: + summary: "Region {{ $labels.environment }} input backlog near its bound" + description: "New inputs are being backpressured or rejected." + - alert: EvoRegionInputsDropped + expr: increase(evo_region_dropped_total{reason!="refused"}[10m]) > 0 + labels: {severity: warning} + annotations: + summary: "Region {{ $labels.environment }} accepted inputs that will never be acknowledged" + description: "Inputs were drained, fenced or quarantined ({{ $labels.reason }}); clients must resubmit to the current owner." + - name: evo-region-storage + rules: + - alert: EvoRegionObjectsLow + expr: evo_region_objects_available_bytes < 2 * evo_region_objects_reserve_bytes + for: 5m + labels: {severity: warning} + annotations: + summary: "Region {{ $labels.environment }} object volume is below twice its reserve" + description: "local-path volumes do not enforce their size; the node's disk is running out." + - alert: EvoRegionObjectsFull + expr: evo_region_objects_available_bytes < evo_region_objects_reserve_bytes + for: 1m + labels: {severity: critical} + annotations: + summary: "Region {{ $labels.environment }} object volume is below its reserve" + description: "New checkpoints are refused; the region cannot commit or acknowledge until space is freed." + - name: aevolution-monitoring + rules: + - alert: Watchdog + expr: vector(1) + labels: {severity: none} + annotations: + summary: "Always firing: proves the alert pipeline end to end" + description: "Route to a heartbeat receiver once one exists; its absence then means monitoring is down." + - alert: AlertmanagerDown + expr: up{job="alertmanager"} == 0 + for: 5m + labels: {severity: critical} + annotations: + summary: "Alertmanager is not scrapeable" + description: "Alerts are evaluated but cannot be routed." diff --git a/deploy/kubernetes/monitoring/config/prometheus.yml b/deploy/kubernetes/monitoring/config/prometheus.yml new file mode 100644 index 0000000..92c18cc --- /dev/null +++ b/deploy/kubernetes/monitoring/config/prometheus.yml @@ -0,0 +1,33 @@ +# M8.5 Prometheus configuration (ADR-0072). Static targets: the region operations +# Service of each disposable environment. The scrape target carries the region's +# identity as labels (ADR-0066): the exposition itself has no world/region labels. +global: + scrape_interval: 15s + scrape_timeout: 5s + evaluation_interval: 15s + +rule_files: + - /etc/prometheus/evo-region.rules.yml + +alerting: + alertmanagers: + - static_configs: + - targets: [alertmanager:9093] + +scrape_configs: + - job_name: evo-region + metrics_path: /metrics + # The exposition is at most 8 KiB (ADR-0066); anything larger is refused. + body_size_limit: 64KB + sample_limit: 64 + static_configs: + - targets: [evo-region-ops.evolution-dev.svc:9100] + labels: {environment: dev, world: "100", region: "1"} + - targets: [evo-region-ops.evolution-stage.svc:9100] + labels: {environment: stage, world: "200", region: "1"} + - job_name: prometheus + static_configs: + - targets: [localhost:9090] + - job_name: alertmanager + static_configs: + - targets: [alertmanager:9093] diff --git a/deploy/kubernetes/monitoring/kustomization.yaml b/deploy/kubernetes/monitoring/kustomization.yaml new file mode 100644 index 0000000..754ab32 --- /dev/null +++ b/deploy/kubernetes/monitoring/kustomization.yaml @@ -0,0 +1,20 @@ +# M8.5 monitoring workload (ADR-0072), synced by the aevolution-monitoring Argo +# application. Configuration and rules live in config/ and are unit-tested in CI +# with promtool (tools/ci/monitoring-check.sh). +apiVersion: kustomize.config.k8s.io/v1beta1 +kind: Kustomization +namespace: aevolution-monitoring +resources: + - prometheus.yaml + - alertmanager.yaml +configMapGenerator: + - name: prometheus-config + files: + - config/prometheus.yml + - config/evo-region.rules.yml + - name: alertmanager-config + files: + - config/alertmanager.yml +labels: + - pairs: + app.kubernetes.io/part-of: aevolution diff --git a/deploy/kubernetes/monitoring/prometheus.yaml b/deploy/kubernetes/monitoring/prometheus.yaml new file mode 100644 index 0000000..266d7d3 --- /dev/null +++ b/deploy/kubernetes/monitoring/prometheus.yaml @@ -0,0 +1,115 @@ +# M8.5 Prometheus (ADR-0072), pinned by digest. Static scrape targets (no API access, +# no RBAC); 15 s scrape and evaluation; retention bounded by time and size. +apiVersion: v1 +kind: PersistentVolumeClaim +metadata: + name: prometheus-data +spec: + accessModes: [ReadWriteOnce] + storageClassName: local-path # Metrics are not world data; Delete reclaim is fine. + resources: + requests: + storage: 5Gi +--- +apiVersion: apps/v1 +kind: Deployment +metadata: + name: prometheus + labels: + app.kubernetes.io/name: prometheus +spec: + replicas: 1 + strategy: + type: Recreate # One TSDB writer on an RWO volume. + selector: + matchLabels: + app.kubernetes.io/name: prometheus + template: + metadata: + labels: + app.kubernetes.io/name: prometheus + app.kubernetes.io/part-of: aevolution + spec: + serviceAccountName: aevolution-prometheus + automountServiceAccountToken: false + securityContext: + runAsNonRoot: true + runAsUser: 65534 + runAsGroup: 65534 + fsGroup: 65534 + seccompProfile: {type: RuntimeDefault} + containers: + - name: prometheus + image: quay.io/prometheus/prometheus:v3.13.3@sha256:6976aa8a60fec930796ce5772b8d12da7a318a5daa8d40d69c5c7819a05eeed7 + args: + - --config.file=/etc/prometheus/prometheus.yml + - --storage.tsdb.path=/prometheus + - --storage.tsdb.retention.time=15d + - --storage.tsdb.retention.size=3GB + - --web.listen-address=:9090 + ports: + - {name: web, containerPort: 9090} + readinessProbe: + httpGet: {path: /-/ready, port: web} + periodSeconds: 10 + livenessProbe: + httpGet: {path: /-/healthy, port: web} + periodSeconds: 20 + failureThreshold: 3 + resources: + requests: {cpu: 100m, memory: 256Mi} + limits: {memory: 1Gi} + securityContext: + allowPrivilegeEscalation: false + readOnlyRootFilesystem: true + capabilities: {drop: [ALL]} + volumeMounts: + - {name: config, mountPath: /etc/prometheus, readOnly: true} + - {name: data, mountPath: /prometheus} + volumes: + - name: config + configMap: + name: prometheus-config + - name: data + persistentVolumeClaim: + claimName: prometheus-data +--- +apiVersion: v1 +kind: Service +metadata: + name: prometheus +spec: + type: ClusterIP + selector: + app.kubernetes.io/name: prometheus + ports: + - {name: web, port: 9090, targetPort: web} +--- +# Prometheus reaches the region operations ports, Alertmanager and itself only. +apiVersion: networking.k8s.io/v1 +kind: NetworkPolicy +metadata: + name: prometheus +spec: + podSelector: + matchLabels: + app.kubernetes.io/name: prometheus + policyTypes: [Egress] + egress: + - to: + - namespaceSelector: + matchExpressions: + - key: kubernetes.io/metadata.name + operator: In + values: [evolution-dev, evolution-stage] + podSelector: + matchLabels: + app.kubernetes.io/name: evo-region + ports: + - {protocol: TCP, port: 9100} + - to: + - podSelector: + matchLabels: + app.kubernetes.io/name: alertmanager + ports: + - {protocol: TCP, port: 9093} diff --git a/deploy/kubernetes/monitoring/tests/evo-region.rules.test.yml b/deploy/kubernetes/monitoring/tests/evo-region.rules.test.yml new file mode 100644 index 0000000..e17849a --- /dev/null +++ b/deploy/kubernetes/monitoring/tests/evo-region.rules.test.yml @@ -0,0 +1,240 @@ +# M8.5 alert injection tests (ADR-0072): promtool test rules. Each alert is shown +# firing on an injected series and staying silent on the healthy or idle case. +rule_files: + - ../config/evo-region.rules.yml +evaluation_interval: 15s +tests: + # Healthy idle region: ready, renewing, idle (debt accrues with no inputs). + - interval: 15s + input_series: + - series: 'up{job="evo-region",environment="dev"}' + values: '1x40' + - series: 'evo_region_ready{environment="dev"}' + values: '1x40' + - series: 'evo_region_phase{environment="dev",phase="running"}' + values: '1x40' + - series: 'evo_region_phase{environment="dev",phase="quarantined"}' + values: '0x40' + - series: 'evo_region_lease_remaining_milliseconds{environment="dev"}' + values: '18000x40' + - series: 'evo_region_tick_debt{environment="dev"}' + values: '0+20x40' + - series: 'evo_region_queued_inputs{environment="dev"}' + values: '0x40' + - series: 'evo_region_queue_capacity_inputs{environment="dev"}' + values: '64x40' + - series: 'evo_region_checkpoint_age_milliseconds{environment="dev"}' + values: '0+15000x40' + - series: 'evo_region_commit_lag_ticks{environment="dev"}' + values: '0x40' + - series: 'evo_region_fences_total{environment="dev"}' + values: '0x40' + - series: 'evo_region_store_failures_total{environment="dev"}' + values: '0x40' + - series: 'evo_region_objects_available_bytes{environment="dev"}' + values: '40000000000x40' + - series: 'evo_region_objects_reserve_bytes{environment="dev"}' + values: '1073741824x40' + - series: 'up{job="alertmanager"}' + values: '1x40' + alert_rule_test: + - {eval_time: 10m, alertname: EvoRegionDown, exp_alerts: []} + - {eval_time: 10m, alertname: EvoRegionNotReady, exp_alerts: []} + - {eval_time: 10m, alertname: EvoRegionQuarantined, exp_alerts: []} + - {eval_time: 10m, alertname: EvoRegionLeaseMargin, exp_alerts: []} + - {eval_time: 10m, alertname: EvoRegionFenced, exp_alerts: []} + - {eval_time: 10m, alertname: EvoRegionStoreFailures, exp_alerts: []} + - {eval_time: 10m, alertname: EvoRegionTickDebt, exp_alerts: []} + - {eval_time: 10m, alertname: EvoRegionCheckpointStale, exp_alerts: []} + - {eval_time: 10m, alertname: EvoRegionQueueSaturated, exp_alerts: []} + - {eval_time: 10m, alertname: EvoRegionObjectsLow, exp_alerts: []} + - {eval_time: 10m, alertname: EvoRegionObjectsFull, exp_alerts: []} + - {eval_time: 10m, alertname: AlertmanagerDown, exp_alerts: []} + - eval_time: 1m + alertname: Watchdog + exp_alerts: + - exp_labels: {severity: none} + exp_annotations: + summary: "Always firing: proves the alert pipeline end to end" + description: "Route to a heartbeat receiver once one exists; its absence then means monitoring is down." + + # Outage: the region disappears, then comes back not ready and quarantined. + - interval: 15s + input_series: + - series: 'up{job="evo-region",environment="dev",world="100",region="1"}' + values: '0x20' + - series: 'evo_region_ready{environment="stage"}' + values: '0x20' + - series: 'evo_region_phase{environment="stage",phase="quarantined"}' + values: '1x20' + alert_rule_test: + - eval_time: 3m + alertname: EvoRegionDown + exp_alerts: + - exp_labels: {severity: critical, job: evo-region, environment: dev, world: "100", region: "1"} + exp_annotations: + summary: "Region dev/100/1 is not scrapeable" + description: "No operations endpoint answers. The process or its pod is gone; a forced deletion recovers in about 20 s, so 2 minutes is a real outage." + - eval_time: 1m + alertname: EvoRegionDown + exp_alerts: [] + - eval_time: 3m + alertname: EvoRegionNotReady + exp_alerts: + - exp_labels: {severity: critical, environment: stage} + exp_annotations: + summary: "Region stage has no authoritative host" + description: "The process is alive but holds no lease or has not verified its head. Check the phase metric and the host log." + - eval_time: 2m + alertname: EvoRegionQuarantined + exp_alerts: + - exp_labels: {severity: critical, environment: stage, phase: quarantined} + exp_annotations: + summary: "Region stage is quarantined" + description: "Committed state or history failed verification. The region stays unavailable; never re-provision it. Audit with evo-world-clone audit." + + # A 20 s lease-bound forced-deletion recovery must not page anyone. + - interval: 15s + input_series: + - series: 'evo_region_ready{environment="dev"}' + values: '1 1 0 0 1 1 1 1 1 1' + alert_rule_test: + - {eval_time: 45s, alertname: EvoRegionNotReady, exp_alerts: []} # During the dip. + - {eval_time: 2m, alertname: EvoRegionNotReady, exp_alerts: []} + + # Lease, fences, storage and database failures. + - interval: 15s + input_series: + - series: 'evo_region_ready{environment="dev"}' + values: '1x20' + - series: 'evo_region_lease_remaining_milliseconds{environment="dev"}' + values: '6000x20' + - series: 'evo_region_fences_total{environment="dev"}' + values: '0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1' + - series: 'evo_region_store_failures_total{environment="dev"}' + values: '0+1x20' + - series: 'evo_region_unknown_outcomes_total{environment="dev"}' + values: '0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1' + alert_rule_test: + - eval_time: 2m + alertname: EvoRegionLeaseMargin + exp_alerts: + - exp_labels: {severity: warning, environment: dev} + exp_annotations: + summary: "Region dev lease is not being renewed" + description: "Healthy renewals keep 15-20 s of lease. Under 10 s for a minute means renewals are failing; the host will fence itself." + - eval_time: 2m + alertname: EvoRegionFenced + exp_alerts: + - exp_labels: {severity: warning, environment: dev} + exp_annotations: + summary: "Region dev host lost authority" + description: "A host was fenced within the last 10 minutes. Expected once after a forced deletion's predecessor resumes; repeated fences point to database or clock trouble." + - eval_time: 3m + alertname: EvoRegionStoreFailures + exp_alerts: + - exp_labels: {severity: warning, environment: dev} + exp_annotations: + summary: "Region dev storage or database failures" + description: "Publication or database calls are failing; progress pauses and nothing is acknowledged until they succeed." + - eval_time: 2m + alertname: EvoRegionUnknownOutcomes + exp_alerts: + - exp_labels: {severity: warning, environment: dev} + exp_annotations: + summary: "Region dev had commits with unknown outcomes" + description: "Commits timed out or lost their connection and were resolved by exact bytes. Occasional is safe; frequent indicates database latency near the 5 s commit deadline." + + # Work waiting: debt, staleness, lag, a saturated queue and dropped inputs. + - interval: 15s + input_series: + - series: 'evo_region_tick_debt{environment="dev"}' + values: '20x20' + - series: 'evo_region_queued_inputs{environment="dev"}' + values: '60x20' + - series: 'evo_region_queue_capacity_inputs{environment="dev"}' + values: '64x20' + - series: 'evo_region_checkpoint_age_milliseconds{environment="dev"}' + values: '400000+15000x20' + - series: 'evo_region_commit_lag_ticks{environment="dev"}' + values: '300x20' + - series: 'evo_region_dropped_total{environment="dev",reason="fenced"}' + values: '0 0 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3' + - series: 'evo_region_dropped_total{environment="dev",reason="refused"}' + values: '0+1x20' + alert_rule_test: + - eval_time: 3m + alertname: EvoRegionTickDebt + exp_alerts: + - exp_labels: {severity: warning, environment: dev} + exp_annotations: + summary: "Region dev is behind real time with inputs waiting" + description: "Due ticks are not being simulated although inputs are queued (catch-up is 8 ticks per step)." + - eval_time: 2m + alertname: EvoRegionCheckpointStale + exp_alerts: + - exp_labels: {severity: warning, environment: dev} + exp_annotations: + summary: "Region dev has not committed for 5 minutes with work pending" + description: "An idle region does not commit; this fires only while inputs or simulated ticks wait for durability." + - eval_time: 3m + alertname: EvoRegionCommitLag + exp_alerts: + - exp_labels: {severity: warning, environment: dev} + exp_annotations: + summary: "Region dev is not committing what it simulates" + description: "More than four full batches are simulated but not durable; acknowledgements are waiting." + - eval_time: 3m + alertname: EvoRegionQueueSaturated + exp_alerts: + - exp_labels: {severity: warning, environment: dev} + exp_annotations: + summary: "Region dev input backlog near its bound" + description: "New inputs are being backpressured or rejected." + - eval_time: 2m + alertname: EvoRegionInputsDropped + exp_alerts: + - exp_labels: {severity: warning, environment: dev, reason: fenced} + exp_annotations: + summary: "Region dev accepted inputs that will never be acknowledged" + description: "Inputs were drained, fenced or quarantined (fenced); clients must resubmit to the current owner." + + # Object volume filling up, then below its reserve. + - interval: 15s + input_series: + - series: 'evo_region_objects_available_bytes{environment="dev"}' + values: '1500000000x30' + - series: 'evo_region_objects_available_bytes{environment="stage"}' + values: '900000000x30' # Below the reserve but above half of it. + - series: 'evo_region_objects_reserve_bytes{environment="dev"}' + values: '1073741824x30' + - series: 'evo_region_objects_reserve_bytes{environment="stage"}' + values: '1073741824x30' + - series: 'up{job="alertmanager"}' + values: '0x30' + alert_rule_test: + - eval_time: 6m + alertname: EvoRegionObjectsLow + exp_alerts: + - exp_labels: {severity: warning, environment: dev} + exp_annotations: + summary: "Region dev object volume is below twice its reserve" + description: "local-path volumes do not enforce their size; the node's disk is running out." + - exp_labels: {severity: warning, environment: stage} + exp_annotations: + summary: "Region stage object volume is below twice its reserve" + description: "local-path volumes do not enforce their size; the node's disk is running out." + - eval_time: 2m + alertname: EvoRegionObjectsFull + exp_alerts: + - exp_labels: {severity: critical, environment: stage} + exp_annotations: + summary: "Region stage object volume is below its reserve" + description: "New checkpoints are refused; the region cannot commit or acknowledge until space is freed." + - eval_time: 6m + alertname: AlertmanagerDown + exp_alerts: + - exp_labels: {severity: critical, job: alertmanager} + exp_annotations: + summary: "Alertmanager is not scrapeable" + description: "Alerts are evaluated but cannot be routed." diff --git a/deploy/kubernetes/platform/cluster/kustomization.yaml b/deploy/kubernetes/platform/cluster/kustomization.yaml index faed6a2..8ba827d 100644 --- a/deploy/kubernetes/platform/cluster/kustomization.yaml +++ b/deploy/kubernetes/platform/cluster/kustomization.yaml @@ -12,5 +12,6 @@ resources: - ../dev - ../stage - ../world0 + - ../monitoring commonAnnotations: argocd.argoproj.io/sync-options: Prune=false,Delete=false diff --git a/deploy/kubernetes/platform/monitoring/guardrails.yaml b/deploy/kubernetes/platform/monitoring/guardrails.yaml new file mode 100644 index 0000000..e254e20 --- /dev/null +++ b/deploy/kubernetes/platform/monitoring/guardrails.yaml @@ -0,0 +1,65 @@ +apiVersion: v1 +kind: ResourceQuota +metadata: + name: aevolution-quota +spec: + hard: + requests.cpu: "1" + requests.memory: 1Gi + limits.memory: 2Gi + pods: "6" + persistentvolumeclaims: "2" + requests.storage: 10Gi + services.loadbalancers: "0" + services.nodeports: "0" +--- +apiVersion: v1 +kind: LimitRange +metadata: + name: aevolution-defaults +spec: + limits: + - type: Container + defaultRequest: {cpu: 50m, memory: 64Mi} + default: {memory: 256Mi} + max: {memory: 2Gi} +--- +apiVersion: networking.k8s.io/v1 +kind: NetworkPolicy +metadata: + name: aevolution-default-deny +spec: + podSelector: {} + policyTypes: [Ingress, Egress] +--- +apiVersion: networking.k8s.io/v1 +kind: NetworkPolicy +metadata: + name: aevolution-allow-dns +spec: + podSelector: {} + policyTypes: [Egress] + egress: + - to: + - namespaceSelector: + matchLabels: + kubernetes.io/metadata.name: kube-system + podSelector: + matchLabels: + k8s-app: kube-dns + ports: + - {protocol: UDP, port: 53} + - {protocol: TCP, port: 53} +--- +# No Kubernetes API access: scrape targets are static, so no token and no RBAC. +apiVersion: v1 +kind: ServiceAccount +metadata: + name: aevolution-prometheus +automountServiceAccountToken: false +--- +apiVersion: v1 +kind: ServiceAccount +metadata: + name: aevolution-alertmanager +automountServiceAccountToken: false diff --git a/deploy/kubernetes/platform/monitoring/kustomization.yaml b/deploy/kubernetes/platform/monitoring/kustomization.yaml new file mode 100644 index 0000000..e5e72f3 --- /dev/null +++ b/deploy/kubernetes/platform/monitoring/kustomization.yaml @@ -0,0 +1,13 @@ +# M8.5 monitoring namespace (ADR-0072): guardrails and identities owned by the +# platform app, like every AEvolution namespace; Prometheus and Alertmanager +# themselves are the aevolution-monitoring workload app. +apiVersion: kustomize.config.k8s.io/v1beta1 +kind: Kustomization +namespace: aevolution-monitoring +resources: + - namespace.yaml + - guardrails.yaml +labels: + - pairs: + app.kubernetes.io/part-of: aevolution + app.kubernetes.io/managed-by: aevolution-platform diff --git a/deploy/kubernetes/platform/monitoring/namespace.yaml b/deploy/kubernetes/platform/monitoring/namespace.yaml new file mode 100644 index 0000000..30b1a79 --- /dev/null +++ b/deploy/kubernetes/platform/monitoring/namespace.yaml @@ -0,0 +1,10 @@ +# Metrics and alerting for the AEvolution environments. Holds no world data. +apiVersion: v1 +kind: Namespace +metadata: + name: aevolution-monitoring + labels: + aevolution.io/environment: monitoring + pod-security.kubernetes.io/enforce: restricted + pod-security.kubernetes.io/enforce-version: v1.32 + pod-security.kubernetes.io/warn: restricted diff --git a/deploy/kubernetes/region/base/kustomization.yaml b/deploy/kubernetes/region/base/kustomization.yaml index 393abb5..3b2a21c 100644 --- a/deploy/kubernetes/region/base/kustomization.yaml +++ b/deploy/kubernetes/region/base/kustomization.yaml @@ -8,6 +8,7 @@ resources: - provision-job.yaml - deployment.yaml - service.yaml + - scrape-policy.yaml labels: - pairs: app.kubernetes.io/part-of: aevolution diff --git a/deploy/kubernetes/region/base/scrape-policy.yaml b/deploy/kubernetes/region/base/scrape-policy.yaml new file mode 100644 index 0000000..9afab4c --- /dev/null +++ b/deploy/kubernetes/region/base/scrape-policy.yaml @@ -0,0 +1,21 @@ +# M8.5: the region's operations port accepts connections only from the AEvolution +# Prometheus. Kubelet probes come from the node and are not affected (Calico). +apiVersion: networking.k8s.io/v1 +kind: NetworkPolicy +metadata: + name: evo-region-scrape +spec: + podSelector: + matchLabels: + app.kubernetes.io/name: evo-region + policyTypes: [Ingress] + ingress: + - from: + - namespaceSelector: + matchLabels: + kubernetes.io/metadata.name: aevolution-monitoring + podSelector: + matchLabels: + app.kubernetes.io/name: prometheus + ports: + - {protocol: TCP, port: 9100} diff --git a/docs/BACKLOG.md b/docs/BACKLOG.md index 7f528ed..57fbbaa 100644 --- a/docs/BACKLOG.md +++ b/docs/BACKLOG.md @@ -10,6 +10,8 @@ names it. Remove an item only when it is done, and record where it was done. | **Backup destination for dev/stage stores.** There are no backups yet (maintainer, 2026-09-29). The plan is to build backup tooling for the AEvolution PostgreSQL VM and each `aevolution-objects` volume up to a pluggable destination, then add the destination. It must be off the host for anything beyond disposable environments. | M8.3, 2026-09-29 | The maintainer will supply a destination later | M8.7 backup and restore; any World0 data | | **Secret manager for Argo-delivered configuration.** Secrets are still created by hand (ADR-0069 amendment). Choose how they reach the cluster through Git or Argo: for example Vault with the Agent Injector (as in the maintainer's other environment), or sealed or external secrets. Until then, rotate by hand. | M8.3, 2026-09-29 | The maintainer wants everything through Argo; a secret manager is not chosen | Stage and World0 credentials; M8.4 | | **Store Job values.** Supply the published registry digests for `aevolution-postgres-tools` and `aevolution-server` (from each CI Publish step's `Digest:` line), and the NOLOGIN runtime group role for dev (`aevolution`) and for stage (`aevolution_stage`). Then include `stores/` in each environment's path. | M8.3, 2026-09-29 | Values pending | M8.3 in-cluster evidence | +| **Alert destination.** Alertmanager routes every alert to a null receiver (ADR-0072). Choose where alerts go (Discord, email, a webhook), and optionally a heartbeat receiver for the always-firing Watchdog. Receiver credentials go in a Secret. | M8.5, 2026-09-30 | The maintainer will decide later | Any claim that alerts reach a person; M8.6 alert evidence; World0 | +| **Kubelet log rotation check.** Confirm the RKE2 nodes' `container-log-max-size` and `container-log-max-files` (default 10Mi × 5), which bound region host logs (ADR-0072). | M8.5, 2026-09-30 | Needs node access | M8.6 | | **Storage after the RAID/disk rework.** Re-run `deploy/kubernetes/inventory/disk-probe-local-path.yaml` (or a new class), and update INVENTORY.md. | M8.1, 2026-09-29 | Maintainer is reorganizing disks | M8.3's final storage decision | ## Accepted risks (maintainer decisions) diff --git a/docs/ROADMAP.md b/docs/ROADMAP.md index f5d424a..f9da858 100644 --- a/docs/ROADMAP.md +++ b/docs/ROADMAP.md @@ -635,3 +635,11 @@ M8.4 implemented and locally qualified (2026-09-29), under ADR-0071: - `region-drill.sh` for rollout and forced-deletion evidence. The in-cluster sync and drill wait on the maintainer. No director yet: assignment is static and lives in Git. + +M8.5 implemented and locally qualified (2026-09-30), under ADR-0072: +- AEvolution-only Prometheus 3.13.3 and Alertmanager 0.34.1, pinned by digest; +- 16 alert rules, tested by injection with promtool in CI; +- object volume capacity metrics; +- the database outbox kept (no broker); +- bounded logs, and a cardinality and cost report. +Alert delivery waits on the maintainer's choice of receiver. diff --git a/docs/STATUS.md b/docs/STATUS.md index a0f8dc5..e8dbe7b 100644 --- a/docs/STATUS.md +++ b/docs/STATUS.md @@ -1064,3 +1064,22 @@ the policy check enforces this. Dev is world 100 and stage world 200, both regio Next: apply `deploy/argocd/dev/project.yaml`, sync, and run `tools/ops/region-drill.sh dev`. + +## M8.5 observability (2026-09-30) + +ADR-0072 adds an AEvolution-only monitoring stack in `aevolution-monitoring`: +- **Components:** Prometheus 3.13.3 and Alertmanager 0.34.1, pinned by digest and + delivered through Argo. +- **Scraping:** static targets, so no Kubernetes API access. The region's port + 9100 is reachable only from Prometheus. +- **Alerts:** 16 rules covering authority, lease, commits, backlog and object + capacity, plus a Watchdog. Each is proven firing and silent by promtool + injection tests in CI. They route to a null receiver until a destination is + chosen. +- **Metrics:** two new capacity gauges (32 fixed series). + +No broker is adopted: the database outbox stays. Logs are bounded by kubelet +rotation. Monitoring costs about 1,300 series, roughly 11 MB a day. + +Next: sync the monitoring app, confirm the targets, and run the `EvoRegionDown` +injection. M8.4's drill results are still to be recorded. diff --git a/docs/adr/0066-operations-endpoints.md b/docs/adr/0066-operations-endpoints.md index 882a9ed..fd1fe5e 100644 --- a/docs/adr/0066-operations-endpoints.md +++ b/docs/adr/0066-operations-endpoints.md @@ -111,3 +111,10 @@ Tests check the per-call budget, including at the connection cap, and that an accepted probe is still answered under replenished pressure. A real server keeps its lease and readiness while a client opens and closes connections continuously through 60 ticks. + +## Amendment: object volume capacity (2026-09-30, M8.5) +Two gauges were added, `evo_region_objects_available_bytes` and +`evo_region_objects_reserve_bytes`, so that ADR-0070's free-space reserve can be +alerted on before new objects are refused (ADR-0072). The fixed exposition is now +32 series. It still has only the `phase` and `reason` labels, and stays under the +8 KiB bound. diff --git a/docs/adr/0072-observability.md b/docs/adr/0072-observability.md new file mode 100644 index 0000000..cc7cefc --- /dev/null +++ b/docs/adr/0072-observability.md @@ -0,0 +1,125 @@ +# ADR-0072: Observability, alerting and background delivery for dev and stage + +## Status +Accepted for M8.5, 2026-09-30. It implements the M8 part of ADR-0015 +(observability), and builds on: +- ADR-0066 (the operations endpoints); +- ADR-0070 (the object reserve); +- ADR-0071 (the region host). + +The maintainer chose an AEvolution-only Prometheus, and left the alert destination +for later. + +## Context +M8.5 asks for: +- tick, lease, commit, storage and backup alerts, and bounded logs; +- an evaluation of the database outbox before any optional JetStream, with no + mandatory Redis or Kafka; +- evidence: alert injection, and a telemetry cardinality and cost report. + +The named risks are broker redelivery changing authority, and logs filling a disk. + +The cluster has no monitoring stack beyond metrics-server (INVENTORY.md). Each +region exposes 32 fixed series. The only labels are a 6-value phase and a 4-value +drop reason. + +## Decision +**Stack.** Namespace `aevolution-monitoring`, with platform-owned guardrails, and +a manually synced Argo application `aevolution-monitoring`. +- **Components.** One Prometheus and one Alertmanager: + - `quay.io/prometheus/prometheus:v3.13.3`, pinned by digest `6976aa8a…`; + - `quay.io/prometheus/alertmanager:v0.34.1`, pinned by digest `e9733baf…`. + + Both are Apache-2.0, and Prometheus is a CNCF graduated project. 3.13 is a + maintained line: 3.13.3 of 2026-09-07 carries current security fixes. +- **Prometheus configuration:** + - Static scrape targets: the `evo-region-ops` Service in dev and stage, labelled + with `environment`, `world` and `region`. No service discovery, so no token + and no RBAC. + - 15 s scrape and evaluation, a 64-sample and 64 KB cap per scrape, and + retention of 15 days or 3 GB. + - A 5 Gi `local-path` volume. Metrics are not world data, so Delete reclaim is + acceptable. +- **Network.** The region namespaces admit port 9100 only from Prometheus. + Prometheus reaches only the region operations ports and Alertmanager; + Alertmanager accepts only Prometheus. Pods run as nobody, with a read-only root + filesystem and no capabilities. +- **Routing.** Alertmanager routes everything to a **null** receiver until a + destination is chosen (docs/BACKLOG.md). Alerts remain visible in its API and + UI. + +**Metrics added.** `evo_region_objects_available_bytes` and +`evo_region_objects_reserve_bytes` make volume capacity alertable before new +objects are refused. The fixed exposition is now 32 series. + +**Alerts** (`config/evo-region.rules.yml`, 16 rules). Thresholds come from +measured behaviour: + +| Area | Alerts | +|---|---| +| Authority | Down (2 min), NotReady (2 min; a forced-deletion recovery takes about 20 s and does not fire), Quarantined, LeaseMargin (below 10 s while ready; healthy is 15–20 s), Fenced | +| Progress | StoreFailures, UnknownOutcomes, CommitLag (over 256 ticks), TickDebt and CheckpointStale (only while inputs or lag wait, because an idle region always accrues debt), QueueSaturated, InputsDropped (not counting `refused`) | +| Storage | ObjectsLow (below 2× the reserve), ObjectsFull (below the reserve) | +| Pipeline | Watchdog (always firing, for a later heartbeat receiver), AlertmanagerDown | + +**Backup alerts** wait for backups (M8.7). There is nothing to alert on until a +backup destination exists; see docs/BACKLOG.md. + +**Alert injection.** `promtool test rules` checks each alert firing on injected +series and staying silent in the healthy, idle and 20 s recovery cases. The CI +`deploy` stage runs it with the `promtool` and `amtool` copied from the same +pinned images, together with syntax checks of the Prometheus and Alertmanager +configurations. + +**Logs.** Container logs are rotated by the kubelet. The defaults are 10 Mi per +file and 5 files per container, 50 Mi at most per container. The actual RKE2 +setting has not been verified on the nodes yet. At most: +- `evo-server` writes one ack line of about 60 bytes per tick, about 100 MB a day + at 20 Hz, so the rotation keeps the most recent 12 hours; +- Prometheus and Alertmanager log little. + +Nothing writes logs to a volume. Durable history lives in PostgreSQL and the +object store, never in logs. + +**Background delivery.** The database outbox stays the delivery mechanism, and no +broker is adopted: +- `region_outbox` is written in the same transaction as each permanent outcome, + delivery is idempotent (`acknowledge_region_outbox`), and delivery never + mutates simulation state (ADR-0063). +- There is no consumer of outcomes yet: no gateway, client or projection. A relay + worker would deliver to nothing, so it is deferred until a consumer exists. +- JetStream is not adopted. There is no demonstrated delivery workload, and a + broker would add a second durability layer to operate and back up. It could + also never hold authority, since its redelivery could only re-announce an + already-final outcome. Redis and Kafka are not needed. + +**Cardinality and cost** (per 15 s scrape): + +| Source | Series | +|---|---| +| Each region | 32, plus about 5 per-target scrape series | +| Two regions (dev, stage) | about 75 | +| Prometheus self-scrape | about 1,000 (typical) | +| Alertmanager | about 200 | +| **Total** | **about 1,300** | + +That is about 87 samples per second. At about 1.5 bytes per compressed sample, +that is about 11 MB a day, or about 170 MB over 15 days, well inside the 3 GB +retention cap and the 5 Gi volume. Each added region adds about 37 series. The +request is 100m CPU and 256 Mi, with a 1 Gi memory limit. Alertmanager requests +50m and 64 Mi, with a 128 Mi limit. + +## Consequences +- Alerts are evaluated and visible, but not delivered until a receiver is chosen. + The Watchdog is ready for a dead-man's-switch receiver. +- Changing a threshold is a Git change that must keep the rule tests passing. +- The kubelet log rotation values should be confirmed on the nodes. +- World0 monitoring (M8.8) reuses the rules with its own target. + +## Alternatives rejected +- **kube-prometheus-stack or the Prometheus Operator.** Cluster-wide CRDs and + RBAC for a single project; not the maintainer's choice. +- **Kubernetes service discovery.** It needs a token and cluster read RBAC for no + gain with two static targets. +- **JetStream, Kafka or Redis for outcomes.** There is no workload yet, and each + would add another stateful system. diff --git a/docs/adr/README.md b/docs/adr/README.md index 1a1f370..95b7cfd 100644 --- a/docs/adr/README.md +++ b/docs/adr/README.md @@ -82,3 +82,4 @@ design. No unadopted candidate is a link dependency of the proof. - [ADR-0069: Environment isolation and World0 protection](0069-environment-isolation.md) - accepted for M8.2 per-environment guardrails, native World0 admission protection, scoped Argo projects and verification. - [ADR-0070: Durable stores for dev and stage](0070-dev-stage-stores.md) - accepted for M8.3 external PostgreSQL per environment, Retain-class node-local object volumes, free-space reserve and in-cluster store qualification. - [ADR-0071: Region host deployment in dev and stage](0071-region-host-deployment.md) - accepted for M8.4 single-replica Recreate host, digest-as-build identity, idempotent provisioning hook and the region drill; no director yet. +- [ADR-0072: Observability, alerting and background delivery](0072-observability.md) - accepted for M8.5 pinned AEvolution Prometheus/Alertmanager, 16 tested alert rules, bounded logs, database outbox over any broker, cardinality/cost report. diff --git a/docs/engineering/VALIDATION.md b/docs/engineering/VALIDATION.md index ab7cc7d..b458ac9 100644 --- a/docs/engineering/VALIDATION.md +++ b/docs/engineering/VALIDATION.md @@ -2811,3 +2811,36 @@ admission policy and the stage and World0 AppProjects, then ran - `region-drill.sh` passed `bash -n`, and refuses World0 before kubectl (also enforced by the policy check). - Not executed: the in-cluster region sync and the drill. + +## M8.5 observability - 2026-09-30 + +- `promtool` 3.13.3 and `amtool` 0.34.1 came from the official release tarballs, + with SHA-256 verified against the published sums: `b349c732…` and `265b9d1e…`. + - The Prometheus configuration syntax is valid, and all 16 rules are valid. + - The alert injection tests passed, as did the Alertmanager `check-config`. +- Rule mutations: five were tried, and three were caught at first. + - **NotReady firing after 15 s:** missed, because the forced-deletion test only + evaluated after recovery. The test now also checks during the dip. + - **ObjectsFull weakened to half the reserve:** missed, because the injected + free space was below both thresholds. The test now uses 900 MB. + - After these fixes all five were caught: the lease threshold, NotReady too + eager, debt counted without inputs, a weakened ObjectsFull, and `refused` + drops counted. +- Windows MSVC CTest 59/59, including the `ops-contracts` update to 32 series. +- WSL Linux `checks` set: 59/59 in each configuration. The format check passed. +- PostgreSQL suites: Release, analysis and ASan/UBSan passed. The ops process + test checks the reserve gauge (1 GiB) and that free space is nonzero. +- `deploy-policy-check.py`: 67 objects, 0 failures. New rules: + - monitoring images must be digest-pinned; + - monitoring pods mount no token; + - no RBAC objects are allowed; + - the monitoring project is scoped. +- Kustomize renders the platform bundle (four namespaces), the monitoring app + (2 Deployments, 2 ConfigMaps, 2 Services, 2 policies, 1 claim), and dev and + stage with the scrape policy. +- The CI `ci-deploy` stage now copies `promtool` and `amtool` from the pinned + images and runs `tools/ci/monitoring-check.sh`. It has not run locally, because + Docker is not running. +- Not executed: anything in the cluster (monitoring sync, target health, the + `EvoRegionDown` injection), alert delivery (no receiver), and the node log + rotation check. diff --git a/docs/infrastructure/ENVIRONMENTS.md b/docs/infrastructure/ENVIRONMENTS.md index 34a4fad..ae44bef 100644 --- a/docs/infrastructure/ENVIRONMENTS.md +++ b/docs/infrastructure/ENVIRONMENTS.md @@ -139,6 +139,27 @@ record: - the first pod's `phase running epoch=... committed=...` line; - that `/readyz` stays 503 while a replacement waits out the lease. +## Monitoring (M8.5, ADR-0072) + +Bootstrap once: `kubectl apply -f deploy/argocd/monitoring/`, then sync +`aevolution-platform` (it adds the `aevolution-monitoring` namespace and its +guardrails) and `aevolution-monitoring`. Re-sync dev and stage, which adds the +`evo-region-scrape` policy. + +```sh +kubectl -n aevolution-monitoring port-forward svc/prometheus 9090 & +curl -s localhost:9090/api/v1/targets | grep -o '"health":"[a-z]*"' # both regions up +curl -s localhost:9090/api/v1/alerts # Watchdog firing only +``` + +**Injection in the cluster** (dev only): +- `kubectl -n evolution-dev scale deploy/evo-region --replicas=0` makes + `EvoRegionDown` pending, then firing after 2 minutes, in Alertmanager (null + receiver). +- Scale back to 1 and it resolves. + +Record the timestamps. The rule tests in CI already prove the thresholds. + ## Break-glass How membership is granted is not decided yet; see docs/BACKLOG.md. Until it is, diff --git a/include/evo/server/ops.hpp b/include/evo/server/ops.hpp index 866b7dd..d2f7755 100644 --- a/include/evo/server/ops.hpp +++ b/include/evo/server/ops.hpp @@ -44,6 +44,9 @@ struct OpsCounters { std::uint64_t checkpoint_age_ms{}; // Since the last confirmed commit or recovery. std::uint64_t acks{}, outcomes{}; std::array dropped{}; // Indexed by Unacknowledged. + // M8.5: object volume free space and the store's reserve (ADR-0070), so capacity + // is alerted on before new objects are refused. + std::uint64_t objects_available_bytes{}, objects_reserve_bytes{}; }; [[nodiscard]] const char* phase_name(HostPhase phase) noexcept; diff --git a/include/evo/server/region_host.hpp b/include/evo/server/region_host.hpp index 3f6d6da..bed1305 100644 --- a/include/evo/server/region_host.hpp +++ b/include/evo/server/region_host.hpp @@ -107,6 +107,10 @@ template class RegionHost { [[nodiscard]] const HostStatus& status() const noexcept { return status_; } + // Read-only, for operational reporting (free space against the reserve). + [[nodiscard]] const persistence::FilesystemObjectStore& objects() const noexcept { + return *objects_; + } // Read-only view for diagnostics/tests; null unless recovered. [[nodiscard]] const State* state() const noexcept { return state_.get(); diff --git a/src/server/ops.cpp b/src/server/ops.cpp index 3d9507c..2a6e063 100644 --- a/src/server/ops.cpp +++ b/src/server/ops.cpp @@ -81,6 +81,10 @@ std::string render_metrics(const HostStatus& s, const OpsCounters& c) { c.queue_capacity_inputs); metric(out, "evo_region_queue_capacity_bytes", "gauge", "Input backlog byte bound", c.queue_capacity_bytes); + metric(out, "evo_region_objects_available_bytes", "gauge", + "Free bytes on the object volume available to the host", c.objects_available_bytes); + metric(out, "evo_region_objects_reserve_bytes", "gauge", + "Free-space reserve below which new objects are refused", c.objects_reserve_bytes); metric(out, "evo_region_commits_total", "counter", "Confirmed commits", s.commits); metric(out, "evo_region_store_failures_total", "counter", "Storage or database failures that paused progress", s.store_failures); diff --git a/tests/ops_tests.cpp b/tests/ops_tests.cpp index 5a26077..accb4e0 100644 --- a/tests/ops_tests.cpp +++ b/tests/ops_tests.cpp @@ -82,8 +82,8 @@ void exposition_is_bounded_and_well_formed() { CHECK(text.find("evo_region_commit_lag_ticks 9\n") != std::string::npos); CHECK(text.find("evo_region_lease_remaining_milliseconds 1234\n") != std::string::npos); CHECK(text.find("evo_region_dropped_total{reason=\"fenced\"} 5\n") != std::string::npos); - // Series count is fixed: 6 phases + 4 reasons + 20 unlabelled families. - CHECK(samples.size() == 6 + 4 + 20); + // Series count is fixed: 6 phases + 4 reasons + 22 unlabelled families. + CHECK(samples.size() == 6 + 4 + 22); // Worst-case values keep the size bounded and never underflow the lag. HostStatus big; @@ -93,7 +93,7 @@ void exposition_is_bounded_and_well_formed() { big.max_tick_debt = big.commits = big.store_failures = big.unknown_outcomes = most; big.fences = big.lease_remaining_ms = most; big.simulated_tick = 1; // Lag saturates at zero rather than wrapping. - OpsCounters bigc{most, most, most, most, most, {most, most, most, most}}; + OpsCounters bigc{most, most, most, most, most, {most, most, most, most}, most, most}; const auto large = render_metrics(big, bigc); CHECK(large.size() <= max_metrics_bytes); CHECK(large.find("evo_region_commit_lag_ticks 0\n") != std::string::npos); diff --git a/tests/region_ops_tests.cpp b/tests/region_ops_tests.cpp index 7d8b5a2..ad024c8 100644 --- a/tests/region_ops_tests.cpp +++ b/tests/region_ops_tests.cpp @@ -97,6 +97,9 @@ void probes_follow_authority_and_stay_bounded() { CHECK(metric(port, "evo_region_lease_remaining_milliseconds") > 0); CHECK(metric(port, "evo_region_queue_capacity_inputs") == 64 && metric(port, "evo_region_queue_capacity_bytes") == 65536); + // M8.5: the default reserve is 1024 MiB, and the volume reports real free space. + CHECK(metric(port, "evo_region_objects_reserve_bytes") == std::uint64_t{1024} * 1024 * 1024 && + metric(port, "evo_region_objects_available_bytes") > 0); auto page = get(port, "/metrics"); CHECK(page.body.find("evo_region_phase{phase=\"running\"} 1\n") != std::string::npos); server.send(input_lines(1, 30, false)); diff --git a/tools/ci/deploy-policy-check.py b/tools/ci/deploy-policy-check.py index 34dde18..2930d52 100644 --- a/tools/ci/deploy-policy-check.py +++ b/tools/ci/deploy-policy-check.py @@ -121,8 +121,8 @@ for path, doc in objects: if doc['kind'] == 'AppProject' and name == 'aevolution-platform': spec = doc['spec'] namespaces = sorted(d.get('namespace') for d in spec.get('destinations', [])) - check(namespaces == ['evolution-dev', 'evolution-stage', 'evolution-world0'], - f'{path}: the platform project targets exactly the three AEvolution namespaces') + check(namespaces == ['aevolution-monitoring', 'evolution-dev', 'evolution-stage', 'evolution-world0'], + f'{path}: the platform project targets exactly the AEvolution namespaces') cluster = sorted((w.get('group'), w.get('kind')) for w in spec.get('clusterResourceWhitelist', [])) check(cluster == [('', 'Namespace'), ('admissionregistration.k8s.io', 'ValidatingAdmissionPolicy'), @@ -131,6 +131,11 @@ for path, doc in objects: f'{path}: the platform project may own only its four cluster kinds') kinds = {w.get('kind') for w in spec.get('namespaceResourceWhitelist', [])} check(not kinds & {'Secret', 'Role', 'RoleBinding', '*'}, f'{path}: platform whitelists a forbidden kind') + elif doc['kind'] == 'AppProject' and name == 'aevolution-monitoring': + spec = doc['spec'] + check([d.get('namespace') for d in spec.get('destinations', [])] == ['aevolution-monitoring'] + and spec.get('clusterResourceWhitelist') == [], + f'{path}: monitoring targets only its namespace, with no cluster kinds') elif doc['kind'] == 'AppProject' and name.startswith('aevolution-'): env = name[len('aevolution-'):] spec = doc['spec'] @@ -161,7 +166,7 @@ for path, doc in objects: check(automated is None, f'{path}: {name} must not auto-sync') check('resources-finalizer.argocd.argoproj.io' not in doc['metadata'].get('finalizers', []), f'{path}: {name} must not cascade-delete') - if project.startswith('aevolution-') and project != 'aevolution-platform': + if project.startswith('aevolution-') and project not in ('aevolution-platform', 'aevolution-monitoring'): check(destination == 'evolution-' + project[len('aevolution-'):], f'{path}: {name} leaves its project environment') check(destination != 'evolution-world0' or name in ('aevolution-world0', 'aevolution-platform'), @@ -222,12 +227,24 @@ for env in ('dev', 'stage'): check(len(server) == 1 and len(builds) == 2 and all('sha256:' + b == server[0] for b in builds), f'{overlay}: the build identity must equal the server image digest') +# M8.5 monitoring (ADR-0072): third-party images are pinned by digest, and Prometheus +# needs no Kubernetes API access. +for path, doc in objects: + if path.startswith('deploy/kubernetes/monitoring/') and doc['kind'] == 'Deployment': + pod = doc['spec']['template']['spec'] + for container in pod['containers']: + check('@sha256:' in container.get('image', ''), f'{path}: {container["name"]} must be pinned by digest') + check(pod.get('automountServiceAccountToken') is False, f'{path}: monitoring pods mount no token') + if path.startswith('deploy/') and doc['kind'] in ('ClusterRole', 'ClusterRoleBinding', 'Role', 'RoleBinding'): + check(False, f'{path}: AEvolution grants no Kubernetes RBAC') + # The platform bundle: every rendered object is protected from prune and deletion. bundle = manifests.get(platform + '/cluster/kustomization.yaml', [{}])[0] check(bundle.get('commonAnnotations', {}).get('argocd.argoproj.io/sync-options') == 'Prune=false,Delete=false', 'the platform bundle must mark every object Prune=false,Delete=false') -check(sorted(bundle.get('resources', [])) == ['../admission', '../dev', '../stage', '../storage', '../world0'], - 'the platform bundle must include admission, storage and the three environments') +check(sorted(bundle.get('resources', [])) == ['../admission', '../dev', '../monitoring', '../stage', '../storage', + '../world0'], + 'the platform bundle must include admission, storage, the environments and monitoring') # Paths a real (non-example) Application syncs must be complete: no REPLACE markers diff --git a/tools/ci/monitoring-check.sh b/tools/ci/monitoring-check.sh new file mode 100644 index 0000000..9fdfce0 --- /dev/null +++ b/tools/ci/monitoring-check.sh @@ -0,0 +1,15 @@ +#!/usr/bin/env bash +# M8.5 monitoring configuration checks (ADR-0072): Prometheus configuration syntax, +# alert rule validity, the alert-injection unit tests, and the Alertmanager routing. +# Needs promtool and amtool on PATH, taken from the same pinned images that run in +# the cluster (the ci-deploy Docker stage copies them). No cluster access. +set -euo pipefail +root="$(cd "$(dirname "$0")/../.." && pwd)" +config="$root/deploy/kubernetes/monitoring/config" +promtool --version | head -1 +amtool --version 2>&1 | head -1 +promtool check config --syntax-only "$config/prometheus.yml" +promtool check rules "$config/evo-region.rules.yml" +(cd "$root/deploy/kubernetes/monitoring/tests" && promtool test rules evo-region.rules.test.yml) +amtool check-config "$config/alertmanager.yml" +echo "monitoring configuration, rules and alert tests passed"