feat: add cluster-infrastructure and api-gateway grafana dashboards

k8s/infra/monitoring/dashboards/api-gateway.yaml

@@ -0,0 +1,36 @@
+apiVersion: v1
+data:
+  api-gateway.json: '{"title":"API Gateway","uid":"api-gateway","schemaVersion":39,"timezone":"browser","time":{"from":"now-6h","to":"now"},"refresh":"30s","tags":["api","gateway","llm"],"panels":[{"id":1,"title":"Gateway
+    Health","type":"row","collapsed":false,"gridPos":{"h":1,"w":24,"x":0,"y":0},"panels":[{"id":2,"title":"Gateway
+    Pods Ready","type":"stat","gridPos":{"h":4,"w":4,"x":0,"y":1},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short","thresholds":{"mode":"absolute","steps":[{"value":null,"color":"red"},{"value":3,"color":"green"}]},"color":{"mode":"thresholds"}}},"targets":[{"expr":"sum(kube_pod_status_ready{namespace=\"api\",condition=\"true\"})"}]},{"id":3,"title":"Probe:
+    healthz","type":"stat","gridPos":{"h":4,"w":4,"x":4,"y":1},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short","mappings":[{"type":"value","options":{"0":{"text":"DOWN","color":"red"},"1":{"text":"UP","color":"green"}}}]}},"targets":[{"expr":"probe_success{instance=~\".*api.riotpiao.com/healthz\"}"}]},{"id":4,"title":"Probe
+    Latency","type":"timeseries","gridPos":{"h":8,"w":8,"x":8,"y":1},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"s"}},"targets":[{"expr":"probe_duration_seconds{instance=~\".*api.riotpiao.com.*\"}","legendFormat":"{{instance}}"}]}]},{"id":10,"title":"Ingress
+    Traffic (nginx)","type":"row","collapsed":true,"gridPos":{"h":1,"w":24,"x":0,"y":1},"panels":[{"id":11,"title":"Request
+    Rate by Status","type":"timeseries","gridPos":{"h":8,"w":8,"x":0,"y":2},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"sum(rate(nginx_ingress_controller_requests{ingress=\"api\"}[5m]))
+    by (status)","legendFormat":"{{status}}"}]},{"id":12,"title":"Error Rate %","type":"timeseries","gridPos":{"h":8,"w":8,"x":8,"y":2},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"percent"}},"targets":[{"expr":"sum(rate(nginx_ingress_controller_requests{ingress=\"api\",status=~\"5..\"}[5m]))
+    / sum(rate(nginx_ingress_controller_requests{ingress=\"api\"}[5m])) * 100","legendFormat":"5xx"},{"expr":"sum(rate(nginx_ingress_controller_requests{ingress=\"api\",status=~\"4..\"}[5m]))
+    / sum(rate(nginx_ingress_controller_requests{ingress=\"api\"}[5m])) * 100","legendFormat":"4xx"}]},{"id":13,"title":"Latency
+    p50/p95/p99","type":"timeseries","gridPos":{"h":8,"w":8,"x":16,"y":2},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"s"}},"targets":[{"expr":"histogram_quantile(0.50,
+    sum(rate(nginx_ingress_controller_request_duration_seconds_bucket{ingress=\"api\"}[5m]))
+    by (le))","legendFormat":"p50"},{"expr":"histogram_quantile(0.95, sum(rate(nginx_ingress_controller_request_duration_seconds_bucket{ingress=\"api\"}[5m]))
+    by (le))","legendFormat":"p95"},{"expr":"histogram_quantile(0.99, sum(rate(nginx_ingress_controller_request_duration_seconds_bucket{ingress=\"api\"}[5m]))
+    by (le))","legendFormat":"p99"}]}]},{"id":20,"title":"LLM Serving","type":"row","collapsed":true,"gridPos":{"h":1,"w":24,"x":0,"y":2},"panels":[{"id":21,"title":"LLM
+    Pods Ready","type":"stat","gridPos":{"h":4,"w":4,"x":0,"y":3},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"sum(kube_pod_status_ready{namespace=\"llm-serving\",condition=\"true\"})"}]},{"id":22,"title":"CPU
+    by Predictor","type":"timeseries","gridPos":{"h":8,"w":8,"x":4,"y":3},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"sum(rate(container_cpu_usage_seconds_total{namespace=\"llm-serving\"}[5m]))
+    by (pod)","legendFormat":"{{pod}}"}]},{"id":23,"title":"Memory by Predictor","type":"timeseries","gridPos":{"h":8,"w":8,"x":12,"y":3},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"bytes"}},"targets":[{"expr":"sum(container_memory_working_set_bytes{namespace=\"llm-serving\"})
+    by (pod)","legendFormat":"{{pod}}"}]},{"id":24,"title":"Predictor Restarts","type":"timeseries","gridPos":{"h":8,"w":4,"x":20,"y":3},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"sum(rate(kube_pod_container_status_restarts_total{namespace=\"llm-serving\"}[15m]))
+    by (pod)","legendFormat":"{{pod}}"}]}]},{"id":30,"title":"Gateway Resources","type":"row","collapsed":true,"gridPos":{"h":1,"w":24,"x":0,"y":3},"panels":[{"id":31,"title":"CPU
+    by Gateway Pod","type":"timeseries","gridPos":{"h":8,"w":8,"x":0,"y":4},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"sum(rate(container_cpu_usage_seconds_total{namespace=\"api\"}[5m]))
+    by (pod)","legendFormat":"{{pod}}"}]},{"id":32,"title":"Memory by Gateway Pod","type":"timeseries","gridPos":{"h":8,"w":8,"x":8,"y":4},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"bytes"}},"targets":[{"expr":"sum(container_memory_working_set_bytes{namespace=\"api\"})
+    by (pod)","legendFormat":"{{pod}}"}]},{"id":33,"title":"Gateway Restarts","type":"timeseries","gridPos":{"h":8,"w":8,"x":16,"y":4},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"sum(rate(kube_pod_container_status_restarts_total{namespace=\"api\"}[15m]))
+    by (pod)","legendFormat":"{{pod}}"}]}]},{"id":40,"title":"Logs","type":"row","collapsed":true,"gridPos":{"h":1,"w":24,"x":0,"y":4},"panels":[{"id":41,"title":"Gateway
+    Logs","type":"logs","gridPos":{"h":10,"w":24,"x":0,"y":5},"datasource":{"type":"loki","uid":"loki"},"targets":[{"expr":"{namespace=\"api\",container=\"gateway\"}"}]},{"id":42,"title":"LLM
+    Serving Logs","type":"logs","gridPos":{"h":10,"w":24,"x":0,"y":15},"datasource":{"type":"loki","uid":"loki"},"targets":[{"expr":"{namespace=\"llm-serving\"}"}]}]}]}'
+kind: ConfigMap
+metadata:
+  annotations:
+    grafana_folder: API
+  labels:
+    grafana_dashboard: '1'
+  name: api-gateway-dashboard
+  namespace: logging

k8s/infra/monitoring/dashboards/cluster-infrastructure.yaml

@@ -0,0 +1,62 @@
+apiVersion: v1
+data:
+  cluster-infrastructure.json: '{"title":"Cluster Infrastructure","uid":"cluster-infra","schemaVersion":39,"timezone":"browser","time":{"from":"now-6h","to":"now"},"refresh":"30s","tags":["infrastructure","k8s"],"panels":[{"id":1,"title":"Cluster
+    Health","type":"row","collapsed":true,"gridPos":{"h":1,"w":24,"x":0,"y":0},"panels":[{"id":2,"title":"Nodes
+    Ready","type":"stat","gridPos":{"h":4,"w":4,"x":0,"y":1},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short","thresholds":{"mode":"absolute","steps":[{"value":null,"color":"red"},{"value":3,"color":"green"}]},"color":{"mode":"thresholds"}}},"targets":[{"expr":"count(kube_node_status_condition{condition=\"Ready\",status=\"true\"}
+    == 1)"}]},{"id":3,"title":"Pods Pending","type":"stat","gridPos":{"h":4,"w":4,"x":4,"y":1},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short","thresholds":{"mode":"absolute","steps":[{"value":null,"color":"green"},{"value":1,"color":"red"}]},"color":{"mode":"thresholds"}}},"targets":[{"expr":"sum(kube_pod_status_phase{phase=\"Pending\"})
+    OR on() vector(0)"}]},{"id":4,"title":"CrashLoopBackOff","type":"stat","gridPos":{"h":4,"w":4,"x":8,"y":1},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short","thresholds":{"mode":"absolute","steps":[{"value":null,"color":"green"},{"value":1,"color":"red"}]},"color":{"mode":"thresholds"}}},"targets":[{"expr":"sum(kube_pod_container_status_waiting_reason{reason=\"CrashLoopBackOff\"})
+    OR on() vector(0)"}]},{"id":5,"title":"OOMKilled (1h)","type":"stat","gridPos":{"h":4,"w":4,"x":12,"y":1},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short","thresholds":{"mode":"absolute","steps":[{"value":null,"color":"green"},{"value":1,"color":"red"}]},"color":{"mode":"thresholds"}}},"targets":[{"expr":"sum(increase(kube_pod_container_status_last_terminated_reason{reason=\"OOMKilled\"}[1h]))
+    OR on() vector(0)"}]},{"id":6,"title":"Deploys Unavailable","type":"stat","gridPos":{"h":4,"w":4,"x":16,"y":1},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short","thresholds":{"mode":"absolute","steps":[{"value":null,"color":"green"},{"value":1,"color":"red"}]},"color":{"mode":"thresholds"}}},"targets":[{"expr":"count(kube_deployment_status_replicas_unavailable
+    > 0) OR on() vector(0)"}]},{"id":7,"title":"Services Down","type":"stat","gridPos":{"h":4,"w":4,"x":20,"y":1},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short","thresholds":{"mode":"absolute","steps":[{"value":null,"color":"green"},{"value":1,"color":"red"}]},"color":{"mode":"thresholds"}}},"targets":[{"expr":"count(probe_success
+    == 0) OR on() vector(0)"}]}]},{"id":10,"title":"Jobs & CronJobs","type":"row","collapsed":true,"gridPos":{"h":1,"w":24,"x":0,"y":1},"panels":[{"id":11,"title":"Failed
+    Jobs","type":"stat","gridPos":{"h":4,"w":4,"x":0,"y":2},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short","thresholds":{"mode":"absolute","steps":[{"value":null,"color":"green"},{"value":1,"color":"red"}]},"color":{"mode":"thresholds"}}},"targets":[{"expr":"count(kube_job_status_failed
+    > 0) OR on() vector(0)"}]},{"id":12,"title":"Failed Jobs Detail","type":"table","gridPos":{"h":8,"w":10,"x":4,"y":2},"datasource":{"type":"prometheus","uid":"prometheus"},"targets":[{"expr":"kube_job_status_failed
+    > 0","format":"table","instant":true}]},{"id":13,"title":"Stuck Jobs (>1h)","type":"table","gridPos":{"h":8,"w":10,"x":14,"y":2},"datasource":{"type":"prometheus","uid":"prometheus"},"targets":[{"expr":"kube_job_status_active
+    == 1 and on(job_name,namespace) (time() - kube_job_status_start_time) > 3600","format":"table","instant":true}]},{"id":14,"title":"CronJob
+    Last Success","type":"timeseries","gridPos":{"h":8,"w":12,"x":0,"y":10},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"dateTimeFromNow"}},"targets":[{"expr":"kube_cronjob_status_last_successful_time{namespace=~\"cicd|kube-system|paperless\"}","legendFormat":"{{namespace}}/{{cronjob}}"}]},{"id":15,"title":"Container
+    Restart Storm (top 10)","type":"timeseries","gridPos":{"h":8,"w":12,"x":12,"y":10},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"topk(10,
+    sum(rate(kube_pod_container_status_restarts_total[15m])) by (namespace, pod))","legendFormat":"{{namespace}}/{{pod}}"}]}]},{"id":20,"title":"Node
+    Resources","type":"row","collapsed":true,"gridPos":{"h":1,"w":24,"x":0,"y":2},"panels":[{"id":21,"title":"CPU
+    % by Node","type":"timeseries","gridPos":{"h":8,"w":8,"x":0,"y":3},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"percent"}},"targets":[{"expr":"(1
+    - avg(rate(node_cpu_seconds_total{mode=\"idle\"}[5m])) by (instance)) * 100","legendFormat":"{{instance}}"}]},{"id":22,"title":"Memory
+    % by Node","type":"timeseries","gridPos":{"h":8,"w":8,"x":8,"y":3},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"percent"}},"targets":[{"expr":"(1
+    - node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes) * 100","legendFormat":"{{instance}}"}]},{"id":23,"title":"Disk
+    % by Node","type":"timeseries","gridPos":{"h":8,"w":8,"x":16,"y":3},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"percent"}},"targets":[{"expr":"(1
+    - node_filesystem_avail_bytes{mountpoint=\"/\"} / node_filesystem_size_bytes{mountpoint=\"/\"})
+    * 100","legendFormat":"{{instance}}"}]},{"id":24,"title":"Load Average","type":"timeseries","gridPos":{"h":8,"w":8,"x":0,"y":11},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"node_load1","legendFormat":"1m
+    {{instance}}"},{"expr":"node_load5","legendFormat":"5m {{instance}}"}]},{"id":25,"title":"Network
+    Errors & Drops","type":"timeseries","gridPos":{"h":8,"w":8,"x":8,"y":11},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"rate(node_network_receive_errs_total[5m])","legendFormat":"rx-err
+    {{instance}}"},{"expr":"rate(node_network_transmit_errs_total[5m])","legendFormat":"tx-err
+    {{instance}}"},{"expr":"rate(node_network_receive_drop_total[5m])","legendFormat":"rx-drop
+    {{instance}}"}]}]},{"id":30,"title":"Control Plane","type":"row","collapsed":true,"gridPos":{"h":1,"w":24,"x":0,"y":3},"panels":[{"id":31,"title":"API
+    Server Up","type":"stat","gridPos":{"h":4,"w":4,"x":0,"y":4},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short","mappings":[{"type":"value","options":{"0":{"text":"DOWN","color":"red"},"1":{"text":"UP","color":"green"}}}]}},"targets":[{"expr":"min(up{job=\"apiserver\"})"}]},{"id":32,"title":"API
+    Server Request Rate","type":"timeseries","gridPos":{"h":8,"w":10,"x":4,"y":4},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"sum(rate(apiserver_request_total[5m]))
+    by (verb, code)","legendFormat":"{{verb}} {{code}}"}]},{"id":33,"title":"API Server
+    Error Rate %","type":"timeseries","gridPos":{"h":8,"w":10,"x":14,"y":4},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"percent"}},"targets":[{"expr":"sum(rate(apiserver_request_total{code=~\"5..\"}[5m]))
+    / sum(rate(apiserver_request_total[5m])) * 100","legendFormat":"5xx %"}]},{"id":34,"title":"API
+    Server Latency","type":"timeseries","gridPos":{"h":8,"w":8,"x":0,"y":12},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"s"}},"targets":[{"expr":"histogram_quantile(0.95,
+    sum(rate(apiserver_request_duration_seconds_bucket[5m])) by (le))","legendFormat":"p95"},{"expr":"histogram_quantile(0.99,
+    sum(rate(apiserver_request_duration_seconds_bucket[5m])) by (le))","legendFormat":"p99"}]},{"id":35,"title":"etcd
+    Request Duration","type":"timeseries","gridPos":{"h":8,"w":8,"x":8,"y":12},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"s"}},"targets":[{"expr":"histogram_quantile(0.99,
+    sum(rate(etcd_request_duration_seconds_bucket[5m])) by (le))","legendFormat":"p99"}]}]},{"id":40,"title":"Storage","type":"row","collapsed":true,"gridPos":{"h":1,"w":24,"x":0,"y":4},"panels":[{"id":41,"title":"Longhorn
+    Disk Capacity","type":"timeseries","gridPos":{"h":8,"w":8,"x":0,"y":5},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"bytes"}},"targets":[{"expr":"longhorn_disk_capacity_bytes","legendFormat":"capacity
+    {{node}}"},{"expr":"longhorn_disk_reservation_bytes","legendFormat":"reserved
+    {{node}}"}]},{"id":42,"title":"PVC Phase","type":"timeseries","gridPos":{"h":8,"w":8,"x":8,"y":5},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"kube_persistentvolumeclaim_status_phase","legendFormat":"{{namespace}}/{{persistentvolumeclaim}}
+    {{phase}}"}]}]},{"id":50,"title":"DNS & Networking","type":"row","collapsed":true,"gridPos":{"h":1,"w":24,"x":0,"y":5},"panels":[{"id":51,"title":"CoreDNS
+    Cache Hit Rate","type":"timeseries","gridPos":{"h":8,"w":8,"x":0,"y":6},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"percentunit"}},"targets":[{"expr":"rate(coredns_cache_hits_total[5m])
+    / (rate(coredns_cache_hits_total[5m]) + rate(coredns_cache_misses_total[5m]))","legendFormat":"{{server}}"}]},{"id":52,"title":"CoreDNS
+    Errors","type":"timeseries","gridPos":{"h":8,"w":8,"x":8,"y":6},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"sum(rate(coredns_dns_responses_total{rcode=~\"SERVFAIL|NXDOMAIN\"}[5m]))
+    by (rcode)","legendFormat":"{{rcode}}"}]}]},{"id":60,"title":"Logs","type":"row","collapsed":true,"gridPos":{"h":1,"w":24,"x":0,"y":6},"panels":[{"id":61,"title":"Error
+    Rate by Namespace","type":"timeseries","gridPos":{"h":8,"w":8,"x":0,"y":7},"datasource":{"type":"prometheus","uid":"prometheus"},"fieldConfig":{"defaults":{"unit":"short"}},"targets":[{"expr":"sum
+    by (namespace) (count_over_time({namespace=~\"kube-system|cert-manager|ingress-nginx|longhorn-system\"}
+    |= \"error\" [5m]))","legendFormat":"{{namespace}}"}]},{"id":62,"title":"Control
+    Plane Logs","type":"logs","gridPos":{"h":10,"w":24,"x":0,"y":15},"datasource":{"type":"loki","uid":"loki"},"targets":[{"expr":"{namespace=\"kube-system\"}"}]},{"id":63,"title":"Cluster
+    Addon Logs","type":"logs","gridPos":{"h":10,"w":24,"x":0,"y":25},"datasource":{"type":"loki","uid":"loki"},"targets":[{"expr":"{namespace=~\"cert-manager|ingress-nginx|longhorn-system\"}"}]}]}]}'
+kind: ConfigMap
+metadata:
+  annotations:
+    grafana_folder: Infrastructure
+  labels:
+    grafana_dashboard: '1'
+  name: cluster-infrastructure-dashboard
+  namespace: logging

k8s/infra/monitoring/dashboards/generate-dashboards.py

@@ -0,0 +1,280 @@
+#!/usr/bin/env python3
+"""Generate consolidated Grafana dashboards as k8s ConfigMap YAML files."""
+
+import json
+import os
+
+DASHBOARD_DIR = os.path.expanduser("~/workplace/homelab/k8s/infra/monitoring/dashboards")
+
+DS_PROM = {"type": "prometheus", "uid": "prometheus"}
+DS_LOKI = {"type": "loki", "uid": "loki"}
+
+
+def stat_panel(id, title, expr, x, y, w=4, h=4, unit="short", mappings=None, thresholds=None):
+    p = {
+        "id": id, "title": title, "type": "stat",
+        "gridPos": {"h": h, "w": w, "x": x, "y": y},
+        "datasource": DS_PROM,
+        "fieldConfig": {"defaults": {"unit": unit}},
+        "targets": [{"expr": expr}],
+    }
+    if mappings:
+        p["fieldConfig"]["defaults"]["mappings"] = mappings
+    if thresholds:
+        p["fieldConfig"]["defaults"]["thresholds"] = thresholds
+        p["fieldConfig"]["defaults"]["color"] = {"mode": "thresholds"}
+    return p
+
+
+def ts_panel(id, title, exprs, x, y, w=8, h=8, unit="short"):
+    targets = []
+    for e in exprs:
+        if isinstance(e, tuple):
+            targets.append({"expr": e[0], "legendFormat": e[1]})
+        else:
+            targets.append({"expr": e, "legendFormat": "{{pod}}"})
+    return {
+        "id": id, "title": title, "type": "timeseries",
+        "gridPos": {"h": h, "w": w, "x": x, "y": y},
+        "datasource": DS_PROM,
+        "fieldConfig": {"defaults": {"unit": unit}},
+        "targets": targets,
+    }
+
+
+def table_panel(id, title, expr, x, y, w=12, h=8):
+    return {
+        "id": id, "title": title, "type": "table",
+        "gridPos": {"h": h, "w": w, "x": x, "y": y},
+        "datasource": DS_PROM,
+        "targets": [{"expr": expr, "format": "table", "instant": True}],
+    }
+
+
+def log_panel(id, title, query, x, y, w=24, h=10):
+    return {
+        "id": id, "title": title, "type": "logs",
+        "gridPos": {"h": h, "w": w, "x": x, "y": y},
+        "datasource": DS_LOKI,
+        "targets": [{"expr": query}],
+    }
+
+
+def row(id, title, y, panels, collapsed=True):
+    return {
+        "id": id, "title": title, "type": "row",
+        "collapsed": collapsed, "gridPos": {"h": 1, "w": 24, "x": 0, "y": y},
+        "panels": panels,
+    }
+
+
+def write_dashboard(filename, dashboard, folder):
+    cm = {
+        "apiVersion": "v1",
+        "kind": "ConfigMap",
+        "metadata": {
+            "name": filename.replace(".yaml", "-dashboard"),
+            "namespace": "logging",
+            "labels": {"grafana_dashboard": "1"},
+            "annotations": {"grafana_folder": folder},
+        },
+        "data": {
+            filename.replace(".yaml", ".json"): json.dumps(dashboard, separators=(",", ":"))
+        },
+    }
+
+    import yaml
+    path = os.path.join(DASHBOARD_DIR, filename)
+    with open(path, "w") as f:
+        yaml.dump(cm, f, default_flow_style=False, allow_unicode=True)
+    print(f"  wrote {path}")
+
+
+# ============================================================================
+# Dashboard 1: Cluster Infrastructure
+# ============================================================================
+
+def build_cluster_infrastructure():
+    zero_thresholds = {"mode": "absolute", "steps": [
+        {"value": None, "color": "green"}, {"value": 1, "color": "red"}
+    ]}
+
+    panels = [
+        row(1, "Cluster Health", 0, [
+            stat_panel(2, "Nodes Ready", 'count(kube_node_status_condition{condition="Ready",status="true"} == 1)', 0, 1, thresholds={"mode":"absolute","steps":[{"value":None,"color":"red"},{"value":3,"color":"green"}]}),
+            stat_panel(3, "Pods Pending", 'sum(kube_pod_status_phase{phase="Pending"}) OR on() vector(0)', 4, 1, thresholds=zero_thresholds),
+            stat_panel(4, "CrashLoopBackOff", 'sum(kube_pod_container_status_waiting_reason{reason="CrashLoopBackOff"}) OR on() vector(0)', 8, 1, thresholds=zero_thresholds),
+            stat_panel(5, "OOMKilled (1h)", 'sum(increase(kube_pod_container_status_last_terminated_reason{reason="OOMKilled"}[1h])) OR on() vector(0)', 12, 1, thresholds=zero_thresholds),
+            stat_panel(6, "Deploys Unavailable", 'count(kube_deployment_status_replicas_unavailable > 0) OR on() vector(0)', 16, 1, thresholds=zero_thresholds),
+            stat_panel(7, "Services Down", 'count(probe_success == 0) OR on() vector(0)', 20, 1, thresholds=zero_thresholds),
+        ]),
+        row(10, "Jobs & CronJobs", 1, [
+            stat_panel(11, "Failed Jobs", 'count(kube_job_status_failed > 0) OR on() vector(0)', 0, 2, thresholds=zero_thresholds),
+            table_panel(12, "Failed Jobs Detail", 'kube_job_status_failed > 0', 4, 2, w=10),
+            table_panel(13, "Stuck Jobs (>1h)", 'kube_job_status_active == 1 and on(job_name,namespace) (time() - kube_job_status_start_time) > 3600', 14, 2, w=10),
+            ts_panel(14, "CronJob Last Success", [
+                ('kube_cronjob_status_last_successful_time{namespace=~"cicd|kube-system|paperless"}', "{{namespace}}/{{cronjob}}")
+            ], 0, 10, w=12, unit="dateTimeFromNow"),
+            ts_panel(15, "Container Restart Storm (top 10)", [
+                ('topk(10, sum(rate(kube_pod_container_status_restarts_total[15m])) by (namespace, pod))', "{{namespace}}/{{pod}}")
+            ], 12, 10, w=12),
+        ]),
+        row(20, "Node Resources", 2, [
+            ts_panel(21, "CPU % by Node", [
+                ('(1 - avg(rate(node_cpu_seconds_total{mode="idle"}[5m])) by (instance)) * 100', "{{instance}}")
+            ], 0, 3, unit="percent"),
+            ts_panel(22, "Memory % by Node", [
+                ('(1 - node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes) * 100', "{{instance}}")
+            ], 8, 3, unit="percent"),
+            ts_panel(23, "Disk % by Node", [
+                ('(1 - node_filesystem_avail_bytes{mountpoint="/"} / node_filesystem_size_bytes{mountpoint="/"}) * 100', "{{instance}}")
+            ], 16, 3, unit="percent"),
+            ts_panel(24, "Load Average", [
+                ("node_load1", "1m {{instance}}"),
+                ("node_load5", "5m {{instance}}"),
+            ], 0, 11),
+            ts_panel(25, "Network Errors & Drops", [
+                ("rate(node_network_receive_errs_total[5m])", "rx-err {{instance}}"),
+                ("rate(node_network_transmit_errs_total[5m])", "tx-err {{instance}}"),
+                ("rate(node_network_receive_drop_total[5m])", "rx-drop {{instance}}"),
+            ], 8, 11),
+        ]),
+        row(30, "Control Plane", 3, [
+            stat_panel(31, "API Server Up", 'min(up{job="apiserver"})', 0, 4, mappings=[
+                {"type":"value","options":{"0":{"text":"DOWN","color":"red"},"1":{"text":"UP","color":"green"}}}
+            ]),
+            ts_panel(32, "API Server Request Rate", [
+                ('sum(rate(apiserver_request_total[5m])) by (verb, code)', "{{verb}} {{code}}")
+            ], 4, 4, w=10),
+            ts_panel(33, "API Server Error Rate %", [
+                ('sum(rate(apiserver_request_total{code=~"5.."}[5m])) / sum(rate(apiserver_request_total[5m])) * 100', "5xx %")
+            ], 14, 4, w=10, unit="percent"),
+            ts_panel(34, "API Server Latency", [
+                ('histogram_quantile(0.95, sum(rate(apiserver_request_duration_seconds_bucket[5m])) by (le))', "p95"),
+                ('histogram_quantile(0.99, sum(rate(apiserver_request_duration_seconds_bucket[5m])) by (le))', "p99"),
+            ], 0, 12, unit="s"),
+            ts_panel(35, "etcd Request Duration", [
+                ('histogram_quantile(0.99, sum(rate(etcd_request_duration_seconds_bucket[5m])) by (le))', "p99"),
+            ], 8, 12, unit="s"),
+        ]),
+        row(40, "Storage", 4, [
+            ts_panel(41, "Longhorn Disk Capacity", [
+                ("longhorn_disk_capacity_bytes", "capacity {{node}}"),
+                ("longhorn_disk_reservation_bytes", "reserved {{node}}"),
+            ], 0, 5, unit="bytes"),
+            ts_panel(42, "PVC Phase", [
+                ('kube_persistentvolumeclaim_status_phase', "{{namespace}}/{{persistentvolumeclaim}} {{phase}}")
+            ], 8, 5),
+        ]),
+        row(50, "DNS & Networking", 5, [
+            ts_panel(51, "CoreDNS Cache Hit Rate", [
+                ('rate(coredns_cache_hits_total[5m]) / (rate(coredns_cache_hits_total[5m]) + rate(coredns_cache_misses_total[5m]))', "{{server}}")
+            ], 0, 6, unit="percentunit"),
+            ts_panel(52, "CoreDNS Errors", [
+                ('sum(rate(coredns_dns_responses_total{rcode=~"SERVFAIL|NXDOMAIN"}[5m])) by (rcode)', "{{rcode}}")
+            ], 8, 6),
+        ]),
+        row(60, "Logs", 6, [
+            ts_panel(61, "Error Rate by Namespace", [
+                ('sum by (namespace) (count_over_time({namespace=~"kube-system|cert-manager|ingress-nginx|longhorn-system"} |= "error" [5m]))', "{{namespace}}")
+            ], 0, 7),
+            log_panel(62, "Control Plane Logs", '{namespace="kube-system"}', 0, 15),
+            log_panel(63, "Cluster Addon Logs", '{namespace=~"cert-manager|ingress-nginx|longhorn-system"}', 0, 25),
+        ]),
+    ]
+
+    return {
+        "title": "Cluster Infrastructure",
+        "uid": "cluster-infra",
+        "schemaVersion": 39,
+        "timezone": "browser",
+        "time": {"from": "now-6h", "to": "now"},
+        "refresh": "30s",
+        "tags": ["infrastructure", "k8s"],
+        "panels": panels,
+    }
+
+
+# ============================================================================
+# Dashboard 3: API Gateway
+# ============================================================================
+
+def build_api_gateway():
+    panels = [
+        row(1, "Gateway Health", 0, [
+            stat_panel(2, "Gateway Pods Ready", 'sum(kube_pod_status_ready{namespace="api",condition="true"})', 0, 1, thresholds={"mode":"absolute","steps":[{"value":None,"color":"red"},{"value":3,"color":"green"}]}),
+            stat_panel(3, "Probe: healthz", 'probe_success{instance=~".*api.riotpiao.com/healthz"}', 4, 1, mappings=[
+                {"type":"value","options":{"0":{"text":"DOWN","color":"red"},"1":{"text":"UP","color":"green"}}}
+            ]),
+            ts_panel(4, "Probe Latency", [
+                ('probe_duration_seconds{instance=~".*api.riotpiao.com.*"}', "{{instance}}")
+            ], 8, 1, unit="s"),
+        ], collapsed=False),
+        row(10, "Ingress Traffic (nginx)", 1, [
+            ts_panel(11, "Request Rate by Status", [
+                ('sum(rate(nginx_ingress_controller_requests{ingress="api"}[5m])) by (status)', "{{status}}")
+            ], 0, 2),
+            ts_panel(12, "Error Rate %", [
+                ('sum(rate(nginx_ingress_controller_requests{ingress="api",status=~"5.."}[5m])) / sum(rate(nginx_ingress_controller_requests{ingress="api"}[5m])) * 100', "5xx"),
+                ('sum(rate(nginx_ingress_controller_requests{ingress="api",status=~"4.."}[5m])) / sum(rate(nginx_ingress_controller_requests{ingress="api"}[5m])) * 100', "4xx"),
+            ], 8, 2, unit="percent"),
+            ts_panel(13, "Latency p50/p95/p99", [
+                ('histogram_quantile(0.50, sum(rate(nginx_ingress_controller_request_duration_seconds_bucket{ingress="api"}[5m])) by (le))', "p50"),
+                ('histogram_quantile(0.95, sum(rate(nginx_ingress_controller_request_duration_seconds_bucket{ingress="api"}[5m])) by (le))', "p95"),
+                ('histogram_quantile(0.99, sum(rate(nginx_ingress_controller_request_duration_seconds_bucket{ingress="api"}[5m])) by (le))', "p99"),
+            ], 16, 2, unit="s"),
+        ]),
+        row(20, "LLM Serving", 2, [
+            stat_panel(21, "LLM Pods Ready", 'sum(kube_pod_status_ready{namespace="llm-serving",condition="true"})', 0, 3),
+            ts_panel(22, "CPU by Predictor", [
+                ('sum(rate(container_cpu_usage_seconds_total{namespace="llm-serving"}[5m])) by (pod)', "{{pod}}")
+            ], 4, 3),
+            ts_panel(23, "Memory by Predictor", [
+                ('sum(container_memory_working_set_bytes{namespace="llm-serving"}) by (pod)', "{{pod}}")
+            ], 12, 3, unit="bytes"),
+            ts_panel(24, "Predictor Restarts", [
+                ('sum(rate(kube_pod_container_status_restarts_total{namespace="llm-serving"}[15m])) by (pod)', "{{pod}}")
+            ], 20, 3, w=4),
+        ]),
+        row(30, "Gateway Resources", 3, [
+            ts_panel(31, "CPU by Gateway Pod", [
+                ('sum(rate(container_cpu_usage_seconds_total{namespace="api"}[5m])) by (pod)', "{{pod}}")
+            ], 0, 4),
+            ts_panel(32, "Memory by Gateway Pod", [
+                ('sum(container_memory_working_set_bytes{namespace="api"}) by (pod)', "{{pod}}")
+            ], 8, 4, unit="bytes"),
+            ts_panel(33, "Gateway Restarts", [
+                ('sum(rate(kube_pod_container_status_restarts_total{namespace="api"}[15m])) by (pod)', "{{pod}}")
+            ], 16, 4),
+        ]),
+        row(40, "Logs", 4, [
+            log_panel(41, "Gateway Logs", '{namespace="api",container="gateway"}', 0, 5),
+            log_panel(42, "LLM Serving Logs", '{namespace="llm-serving"}', 0, 15),
+        ]),
+    ]
+
+    return {
+        "title": "API Gateway",
+        "uid": "api-gateway",
+        "schemaVersion": 39,
+        "timezone": "browser",
+        "time": {"from": "now-6h", "to": "now"},
+        "refresh": "30s",
+        "tags": ["api", "gateway", "llm"],
+        "panels": panels,
+    }
+
+
+# ============================================================================
+# Generate
+# ============================================================================
+
+print("Generating dashboards...")
+
+# Dashboard 1
+write_dashboard("cluster-infrastructure.yaml", build_cluster_infrastructure(), "Infrastructure")
+
+# Dashboard 3
+write_dashboard("api-gateway.yaml", build_api_gateway(), "API")
+
+print("Done.")

k8s/infra/monitoring/kustomization.yaml

@@ -29,3 +29,5 @@ resources:
   - dashboards/svc-grafana.yaml
   - dashboards/svc-minio.yaml
   - dashboards/svc-vault.yaml
+  - dashboards/cluster-infrastructure.yaml
+  - dashboards/api-gateway.yaml