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

#!/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.")