uptop/vhs/backfill/main.go

package main

import (
	"database/sql"
	"fmt"
	"math/rand"
	"os"
	"time"

	_ "github.com/mattn/go-sqlite3"
)

func main() {
	if len(os.Args) < 2 {
		fmt.Fprintln(os.Stderr, "usage: backfill <db-path>")
		os.Exit(1)
	}
	db, err := sql.Open("sqlite3", os.Args[1])
	if err != nil {
		fmt.Fprintf(os.Stderr, "open: %v\n", err)
		os.Exit(1)
	}
	defer db.Close()

	ids, err := loadSiteIDs(db)
	if err != nil {
		fmt.Fprintf(os.Stderr, "load site IDs: %v\n", err)
		os.Exit(1)
	}

	rng := rand.New(rand.NewSource(42))
	now := time.Now().UTC()

	if err := backfillHistory(db, rng, now, ids); err != nil {
		fmt.Fprintf(os.Stderr, "history: %v\n", err)
		os.Exit(1)
	}
	if err := backfillStateChanges(db, now, ids); err != nil {
		fmt.Fprintf(os.Stderr, "state changes: %v\n", err)
		os.Exit(1)
	}
	if err := backfillLogs(db, now); err != nil {
		fmt.Fprintf(os.Stderr, "logs: %v\n", err)
		os.Exit(1)
	}
	if err := backfillNodes(db, now); err != nil {
		fmt.Fprintf(os.Stderr, "nodes: %v\n", err)
		os.Exit(1)
	}
	if err := backfillMaintenance(db, now, ids); err != nil {
		fmt.Fprintf(os.Stderr, "maintenance: %v\n", err)
		os.Exit(1)
	}

	var count int
	db.QueryRow("SELECT COUNT(*) FROM check_history").Scan(&count)
	fmt.Printf("Backfill complete: %d check records\n", count)

	var token string
	if err := db.QueryRow("SELECT token FROM sites WHERE name='Nightly Backup'").Scan(&token); err == nil {
		fmt.Printf("PUSH_TOKEN=%s\n", token)
	}
}

func loadSiteIDs(db *sql.DB) (map[string]int, error) {
	rows, err := db.Query("SELECT id, name FROM sites")
	if err != nil {
		return nil, err
	}
	defer rows.Close()
	ids := make(map[string]int)
	for rows.Next() {
		var id int
		var name string
		if err := rows.Scan(&id, &name); err != nil {
			return nil, err
		}
		ids[name] = id
	}
	return ids, rows.Err()
}

type monitorProfile struct {
	name     string
	minMs    int
	maxMs    int
	downFrom int // check index where DOWN starts (-1 = never)
}

func backfillHistory(db *sql.DB, rng *rand.Rand, now time.Time, ids map[string]int) error {
	profiles := []monitorProfile{
		{"Nextcloud", 40, 80, -1},
		{"Jellyfin", 80, 200, -1},
		{"Home Assistant", 15, 45, -1},
		{"Gitea", 40, 90, -1},
		{"Traefik Dashboard", 5, 25, -1},
		{"Vaultwarden", 50, 130, -1},
		{"Personal Blog", 25, 65, -1},
		{"Immich", 100, 280, -1},    // spikes handled below
		{"Auth Portal", 30, 70, 40}, // DOWN after check 40
		{"Edge Router", 5, 15, -1},  // ping
		{"Postgres", 1, 5, -1},      // port
		{"DNS Primary", 10, 30, -1},
	}

	tx, err := db.Begin()
	if err != nil {
		return err
	}
	defer tx.Rollback()

	stmt, err := tx.Prepare("INSERT INTO check_history (site_id, latency_ns, is_up, checked_at) VALUES (?, ?, ?, ?)")
	if err != nil {
		return err
	}
	defer stmt.Close()

	const total = 60
	for _, p := range profiles {
		siteID, ok := ids[p.name]
		if !ok {
			continue
		}
		for i := 0; i < total; i++ {
			minutesAgo := (total - i) * 24
			checkedAt := now.Add(-time.Duration(minutesAgo) * time.Minute)

			var latencyNs int64
			isUp := true

			if p.downFrom >= 0 && i >= p.downFrom {
				latencyNs = 0
				isUp = false
			} else {
				ms := p.minMs + rng.Intn(p.maxMs-p.minMs)
				if p.name == "Immich" && i%17 == 0 {
					ms = 250 + rng.Intn(100)
				}
				latencyNs = int64(ms) * 1_000_000
			}

			if _, err := stmt.Exec(siteID, latencyNs, isUp, checkedAt.Format("2006-01-02 15:04:05")); err != nil {
				return err
			}
		}
	}
	return tx.Commit()
}

func backfillStateChanges(db *sql.DB, now time.Time, ids map[string]int) error {
	type sc struct {
		name   string
		from   string
		to     string
		reason string
		at     time.Time
	}
	changes := []sc{
		{"Nextcloud", "UP", "DOWN", "read timeout", now.Add(-3 * 24 * time.Hour).Add(-5 * time.Minute)},
		{"Nextcloud", "DOWN", "UP", "", now.Add(-3 * 24 * time.Hour)},
		{"Jellyfin", "UP", "DOWN", "connection reset", now.Add(-18 * time.Hour).Add(-3 * time.Minute)},
		{"Jellyfin", "DOWN", "UP", "", now.Add(-18 * time.Hour)},
		{"Auth Portal", "UP", "DOWN", "connection refused", now.Add(-8 * time.Hour)},
		{"Immich", "UP", "DOWN", "502 Bad Gateway", now.Add(-12 * time.Hour).Add(-8 * time.Minute)},
		{"Immich", "DOWN", "UP", "", now.Add(-12 * time.Hour)},
	}

	tx, err := db.Begin()
	if err != nil {
		return err
	}
	defer tx.Rollback()

	stmt, err := tx.Prepare("INSERT INTO state_changes (site_id, from_status, to_status, error_reason, changed_at) VALUES (?, ?, ?, ?, ?)")
	if err != nil {
		return err
	}
	defer stmt.Close()

	for _, c := range changes {
		siteID, ok := ids[c.name]
		if !ok {
			continue
		}
		if _, err := stmt.Exec(siteID, c.from, c.to, c.reason, c.at.Format("2006-01-02 15:04:05")); err != nil {
			return err
		}
	}
	return tx.Commit()
}

func backfillLogs(db *sql.DB, now time.Time) error {
	type logEntry struct {
		msg string
		at  time.Time
	}
	logs := []logEntry{
		{"[06:12] Monitor 'Auth Portal' confirmed DOWN: connection refused", now.Add(-8 * time.Hour)},
		{"[06:12] Monitor 'Auth Portal' failed check 2/2", now.Add(-8*time.Hour - 30*time.Second)},
		{"[06:11] Monitor 'Auth Portal' failed check 1/2", now.Add(-8*time.Hour - 60*time.Second)},
		{"[12:33] Monitor 'Immich' recovered (was down 8m)", now.Add(-12 * time.Hour)},
		{"[12:25] Monitor 'Immich' confirmed DOWN: 502 Bad Gateway", now.Add(-12*time.Hour - 8*time.Minute)},
		{"[12:25] Monitor 'Immich' failed check 3/3", now.Add(-12*time.Hour - 8*time.Minute - 30*time.Second)},
		{"[12:25] Monitor 'Immich' failed check 2/3", now.Add(-12*time.Hour - 8*time.Minute - 60*time.Second)},
		{"[12:24] Monitor 'Immich' failed check 1/3", now.Add(-12*time.Hour - 9*time.Minute)},
		{"[06:14] Monitor 'Jellyfin' recovered (was down 3m)", now.Add(-18 * time.Hour)},
		{"[06:11] Monitor 'Jellyfin' confirmed DOWN: connection reset", now.Add(-18*time.Hour - 3*time.Minute)},
		{"[06:11] Monitor 'Jellyfin' failed check 2/2", now.Add(-18*time.Hour - 3*time.Minute - 30*time.Second)},
		{"[06:10] Monitor 'Jellyfin' failed check 1/2", now.Add(-18*time.Hour - 4*time.Minute)},
		{"[23:45] SSL certificate for 'Personal Blog' expires in 42 days", now.Add(-28 * time.Hour)},
		{"[08:00] Loaded check history from database", now.Add(-32*time.Hour - 30*time.Minute)},
		{"[08:00] Engine RESUMED (Active)", now.Add(-32*time.Hour - 30*time.Minute - 5*time.Second)},
	}

	tx, err := db.Begin()
	if err != nil {
		return err
	}
	defer tx.Rollback()

	stmt, err := tx.Prepare("INSERT INTO logs (message, created_at) VALUES (?, ?)")
	if err != nil {
		return err
	}
	defer stmt.Close()

	for _, l := range logs {
		if _, err := stmt.Exec(l.msg, l.at.Format("2006-01-02 15:04:05")); err != nil {
			return err
		}
	}
	return tx.Commit()
}

func backfillNodes(db *sql.DB, now time.Time) error {
	_, err := db.Exec(
		"INSERT OR REPLACE INTO nodes (id, name, region, last_seen, version) VALUES (?, ?, ?, ?, ?)",
		"node-1", "leader", "us-east", now.Format("2006-01-02 15:04:05"), "2026.05.1",
	)
	return err
}

func backfillMaintenance(db *sql.DB, now time.Time, ids map[string]int) error {
	tx, err := db.Begin()
	if err != nil {
		return err
	}
	defer tx.Rollback()

	stmt, err := tx.Prepare("INSERT INTO maintenance_windows (monitor_id, title, description, type, start_time, end_time, created_by) VALUES (?, ?, ?, ?, ?, ?, ?)")
	if err != nil {
		return err
	}
	defer stmt.Close()

	jellyfinID := ids["Jellyfin"]
	past := now.Add(-3 * 24 * time.Hour)
	if _, err := stmt.Exec(jellyfinID, "Jellyfin upgrade", "Upgrade to v10.10 + plugin updates", "maintenance",
		past.Format("2006-01-02 15:04:05"),
		past.Add(2*time.Hour).Format("2006-01-02 15:04:05"),
		"admin"); err != nil {
		return err
	}

	future := now.Add(2 * 24 * time.Hour)
	if _, err := stmt.Exec(0, "Network switch replacement", "Replacing core switch in rack 2", "maintenance",
		future.Format("2006-01-02 15:04:05"),
		future.Add(4*time.Hour).Format("2006-01-02 15:04:05"),
		"admin"); err != nil {
		return err
	}

	return tx.Commit()
}