diff --git a/misc/api.func b/misc/api.func index d910b9884..7d02a7912 100644 --- a/misc/api.func +++ b/misc/api.func @@ -37,6 +37,10 @@ TELEMETRY_URL="https://telemetry.community-scripts.org/telemetry" # Timeout for telemetry requests (seconds) TELEMETRY_TIMEOUT=5 +# Repository source identifier (auto-transformed by CI on promotion to ProxmoxVE) +# DO NOT CHANGE - this is used by the telemetry service to route data to the correct collection +REPO_SOURCE="community-scripts/ProxmoxVED" + # ============================================================================== # SECTION 1: ERROR CODE DESCRIPTIONS # ============================================================================== @@ -350,7 +354,8 @@ post_to_api() { "gpu_vendor": "${gpu_vendor}", "gpu_model": "${gpu_model}", "gpu_passthrough": "${gpu_passthrough}", - "ram_speed": "${ram_speed}" + "ram_speed": "${ram_speed}", + "repo_source": "${REPO_SOURCE}" } EOF ) @@ -421,7 +426,8 @@ post_to_api_vm() { "os_type": "${var_os:-}", "os_version": "${var_version:-}", "pve_version": "${pve_version}", - "method": "${METHOD:-default}" + "method": "${METHOD:-default}", + "repo_source": "${REPO_SOURCE}" } EOF ) @@ -541,7 +547,8 @@ post_update_to_api() { "gpu_vendor": "${gpu_vendor}", "gpu_model": "${gpu_model}", "gpu_passthrough": "${gpu_passthrough}", - "ram_speed": "${ram_speed}" + "ram_speed": "${ram_speed}", + "repo_source": "${REPO_SOURCE}" } EOF ) @@ -671,7 +678,8 @@ post_tool_to_api() { "error": "${error}", "error_category": "${error_category}", "install_duration": ${duration:-0}, - "pve_version": "${pve_version}" + "pve_version": "${pve_version}", + "repo_source": "${REPO_SOURCE}" } EOF ) @@ -734,7 +742,8 @@ post_addon_to_api() { "error_category": "${error_category}", "install_duration": ${duration:-0}, "os_type": "${os_type}", - "os_version": "${os_version}" + "os_version": "${os_version}", + "repo_source": "${REPO_SOURCE}" } EOF ) @@ -820,7 +829,8 @@ post_update_to_api_extended() { "error_category": "${error_category}", "install_duration": ${duration:-0}, "gpu_vendor": "${gpu_vendor}", - "gpu_passthrough": "${gpu_passthrough}" + "gpu_passthrough": "${gpu_passthrough}", + "repo_source": "${REPO_SOURCE}" } EOF ) diff --git a/misc/data/Dockerfile b/misc/data/Dockerfile index d9228dafe..3d58795c0 100644 --- a/misc/data/Dockerfile +++ b/misc/data/Dockerfile @@ -3,13 +3,13 @@ WORKDIR /src COPY go.mod go.sum* ./ RUN go mod download 2>/dev/null || true COPY . . -RUN go build -trimpath -ldflags "-s -w" -o /out/telemetry-ingest . -RUN go build -trimpath -ldflags "-s -w" -o /out/migrate migrate.go +RUN go build -trimpath -ldflags "-s -w" -o /out/telemetry-service . +RUN go build -trimpath -ldflags "-s -w" -o /out/migrate ./migration/migrate.go FROM alpine:3.23 RUN apk add --no-cache ca-certificates tzdata WORKDIR /app -COPY --from=build /out/telemetry-ingest /app/telemetry-ingest +COPY --from=build /out/telemetry-service /app/telemetry-service COPY --from=build /out/migrate /app/migrate COPY entrypoint.sh /app/entrypoint.sh RUN chmod +x /app/entrypoint.sh /app/migrate diff --git a/misc/data/alerts.go b/misc/data/alerts.go index 198e7a599..01f1a57f1 100644 --- a/misc/data/alerts.go +++ b/misc/data/alerts.go @@ -25,16 +25,54 @@ type AlertConfig struct { FailureThreshold float64 // Alert when failure rate exceeds this (e.g., 20.0 = 20%) CheckInterval time.Duration // How often to check Cooldown time.Duration // Minimum time between alerts + + // Weekly Report settings + WeeklyReportEnabled bool // Enable weekly summary reports + WeeklyReportDay time.Weekday // Day to send report (0=Sunday, 1=Monday, etc.) + WeeklyReportHour int // Hour to send report (0-23) +} + +// WeeklyReportData contains aggregated weekly statistics +type WeeklyReportData struct { + CalendarWeek int + Year int + StartDate time.Time + EndDate time.Time + TotalInstalls int + SuccessCount int + FailedCount int + SuccessRate float64 + TopApps []AppStat + TopFailedApps []AppStat + ComparedToPrev WeekComparison + OsDistribution map[string]int + TypeDistribution map[string]int +} + +// AppStat represents statistics for a single app +type AppStat struct { + Name string + Total int + Failed int + FailureRate float64 +} + +// WeekComparison shows changes compared to previous week +type WeekComparison struct { + InstallsChange int // Difference in total installs + InstallsPercent float64 // Percentage change + FailRateChange float64 // Change in failure rate (percentage points) } // Alerter handles alerting functionality type Alerter struct { - cfg AlertConfig - lastAlertAt time.Time - mu sync.Mutex - pb *PBClient - lastStats alertStats - alertHistory []AlertEvent + cfg AlertConfig + lastAlertAt time.Time + lastWeeklyReport time.Time + mu sync.Mutex + pb *PBClient + lastStats alertStats + alertHistory []AlertEvent } type alertStats struct { @@ -74,6 +112,12 @@ func (a *Alerter) Start() { go a.monitorLoop() log.Printf("INFO: alert monitoring started (threshold: %.1f%%, interval: %v)", a.cfg.FailureThreshold, a.cfg.CheckInterval) + + // Start weekly report scheduler if enabled + if a.cfg.WeeklyReportEnabled { + go a.weeklyReportLoop() + log.Printf("INFO: weekly report scheduler started (day: %s, hour: %02d:00)", a.cfg.WeeklyReportDay, a.cfg.WeeklyReportHour) + } } func (a *Alerter) monitorLoop() { @@ -162,13 +206,21 @@ This is an automated alert from the telemetry service. } func (a *Alerter) sendEmail(subject, body string) error { + return a.sendEmailWithType(subject, body, "text/plain") +} + +func (a *Alerter) sendHTMLEmail(subject, body string) error { + return a.sendEmailWithType(subject, body, "text/html") +} + +func (a *Alerter) sendEmailWithType(subject, body, contentType string) error { // Build message var msg bytes.Buffer msg.WriteString(fmt.Sprintf("From: %s\r\n", a.cfg.SMTPFrom)) msg.WriteString(fmt.Sprintf("To: %s\r\n", strings.Join(a.cfg.SMTPTo, ", "))) msg.WriteString(fmt.Sprintf("Subject: %s\r\n", subject)) msg.WriteString("MIME-Version: 1.0\r\n") - msg.WriteString("Content-Type: text/plain; charset=UTF-8\r\n") + msg.WriteString(fmt.Sprintf("Content-Type: %s; charset=UTF-8\r\n", contentType)) msg.WriteString("\r\n") msg.WriteString(body) @@ -264,4 +316,538 @@ This is an automated test message. // Helper for timeout context func newTimeoutContext(d time.Duration) (context.Context, context.CancelFunc) { return context.WithTimeout(context.Background(), d) +} + +// weeklyReportLoop checks periodically if it's time to send the weekly report +func (a *Alerter) weeklyReportLoop() { + // Check every hour + ticker := time.NewTicker(1 * time.Hour) + defer ticker.Stop() + + for range ticker.C { + a.checkAndSendWeeklyReport() + } +} + +// checkAndSendWeeklyReport sends the weekly report if it's the right time +func (a *Alerter) checkAndSendWeeklyReport() { + now := time.Now() + + // Check if it's the right day and hour + if now.Weekday() != a.cfg.WeeklyReportDay || now.Hour() != a.cfg.WeeklyReportHour { + return + } + + a.mu.Lock() + // Check if we already sent a report this week + _, lastWeek := a.lastWeeklyReport.ISOWeek() + _, currentWeek := now.ISOWeek() + if a.lastWeeklyReport.Year() == now.Year() && lastWeek == currentWeek { + a.mu.Unlock() + return + } + a.mu.Unlock() + + // Send the weekly report + if err := a.SendWeeklyReport(); err != nil { + log.Printf("ERROR: failed to send weekly report: %v", err) + } +} + +// SendWeeklyReport generates and sends the weekly summary email +func (a *Alerter) SendWeeklyReport() error { + if !a.cfg.Enabled || a.cfg.SMTPHost == "" { + return fmt.Errorf("alerting not configured") + } + + ctx, cancel := newTimeoutContext(30 * time.Second) + defer cancel() + + // Get data for the past week + reportData, err := a.fetchWeeklyReportData(ctx) + if err != nil { + return fmt.Errorf("failed to fetch weekly data: %w", err) + } + + // Generate email content + subject := fmt.Sprintf("[ProxmoxVED] Weekly Report - Week %d, %d", reportData.CalendarWeek, reportData.Year) + body := a.generateWeeklyReportHTML(reportData) + + if err := a.sendHTMLEmail(subject, body); err != nil { + return fmt.Errorf("failed to send email: %w", err) + } + + a.mu.Lock() + a.lastWeeklyReport = time.Now() + a.alertHistory = append(a.alertHistory, AlertEvent{ + Timestamp: time.Now(), + Type: "weekly_report", + Message: fmt.Sprintf("Weekly report KW %d/%d sent", reportData.CalendarWeek, reportData.Year), + }) + a.mu.Unlock() + + log.Printf("INFO: weekly report KW %d/%d sent successfully", reportData.CalendarWeek, reportData.Year) + return nil +} + +// fetchWeeklyReportData collects data for the weekly report +func (a *Alerter) fetchWeeklyReportData(ctx context.Context) (*WeeklyReportData, error) { + // Calculate the previous week's date range (Mon-Sun) + now := time.Now() + + // Find last Monday + daysToLastMonday := int(now.Weekday() - time.Monday) + if daysToLastMonday < 0 { + daysToLastMonday += 7 + } + // Go back to the Monday of LAST week + lastMonday := now.AddDate(0, 0, -daysToLastMonday-7) + lastMonday = time.Date(lastMonday.Year(), lastMonday.Month(), lastMonday.Day(), 0, 0, 0, 0, lastMonday.Location()) + lastSunday := lastMonday.AddDate(0, 0, 6) + lastSunday = time.Date(lastSunday.Year(), lastSunday.Month(), lastSunday.Day(), 23, 59, 59, 0, lastSunday.Location()) + + // Get calendar week + year, week := lastMonday.ISOWeek() + + // Fetch current week's data (7 days) + currentData, err := a.pb.FetchDashboardData(ctx, 7) + if err != nil { + return nil, fmt.Errorf("failed to fetch current week data: %w", err) + } + + // Fetch previous week's data for comparison (14 days, we'll compare) + prevData, err := a.pb.FetchDashboardData(ctx, 14) + if err != nil { + // Non-fatal, just log + log.Printf("WARN: could not fetch previous week data: %v", err) + prevData = nil + } + + // Build report data + report := &WeeklyReportData{ + CalendarWeek: week, + Year: year, + StartDate: lastMonday, + EndDate: lastSunday, + TotalInstalls: currentData.TotalInstalls, + SuccessCount: currentData.SuccessCount, + FailedCount: currentData.FailedCount, + OsDistribution: make(map[string]int), + TypeDistribution: make(map[string]int), + } + + // Calculate success rate + if report.TotalInstalls > 0 { + report.SuccessRate = float64(report.SuccessCount) / float64(report.TotalInstalls) * 100 + } + + // Top 5 installed apps + for i, app := range currentData.TopApps { + if i >= 5 { + break + } + report.TopApps = append(report.TopApps, AppStat{ + Name: app.App, + Total: app.Count, + }) + } + + // Top 5 failed apps + for i, app := range currentData.FailedApps { + if i >= 5 { + break + } + report.TopFailedApps = append(report.TopFailedApps, AppStat{ + Name: app.App, + Total: app.TotalCount, + Failed: app.FailedCount, + FailureRate: app.FailureRate, + }) + } + + // OS distribution + for _, os := range currentData.OsDistribution { + report.OsDistribution[os.Os] = os.Count + } + + // Type distribution (LXC vs VM) + for _, t := range currentData.TypeStats { + report.TypeDistribution[t.Type] = t.Count + } + + // Calculate comparison to previous week + if prevData != nil { + // Previous week stats (subtract current from 14-day total) + prevInstalls := prevData.TotalInstalls - currentData.TotalInstalls + prevFailed := prevData.FailedCount - currentData.FailedCount + prevSuccess := prevData.SuccessCount - currentData.SuccessCount + + if prevInstalls > 0 { + prevFailRate := float64(prevFailed) / float64(prevInstalls) * 100 + currentFailRate := 100 - report.SuccessRate + + report.ComparedToPrev.InstallsChange = report.TotalInstalls - prevInstalls + if prevInstalls > 0 { + report.ComparedToPrev.InstallsPercent = float64(report.TotalInstalls-prevInstalls) / float64(prevInstalls) * 100 + } + report.ComparedToPrev.FailRateChange = currentFailRate - prevFailRate + _ = prevSuccess // suppress unused warning + } + } + + return report, nil +} + +// generateWeeklyReportHTML creates the HTML email body for the weekly report +func (a *Alerter) generateWeeklyReportHTML(data *WeeklyReportData) string { + var b strings.Builder + + // HTML Email Template + b.WriteString(` + + + + + + + + +
+ + + + + + + + + + + + + + + + +`) + + // Week comparison + if data.ComparedToPrev.InstallsChange != 0 || data.ComparedToPrev.FailRateChange != 0 { + installIcon := "📈" + installColor := "#16a34a" + if data.ComparedToPrev.InstallsChange < 0 { + installIcon = "📉" + installColor = "#dc2626" + } + failIcon := "✅" + failColor := "#16a34a" + if data.ComparedToPrev.FailRateChange > 0 { + failIcon = "⚠️" + failColor = "#dc2626" + } + + b.WriteString(` + + +`) + } + + // Top 5 Installed Scripts + b.WriteString(` + + +`) + + // Top 5 Failed Scripts + b.WriteString(` + + +`) + + // Type Distribution + if len(data.TypeDistribution) > 0 { + b.WriteString(` + + +`) + } + + // OS Distribution + if len(data.OsDistribution) > 0 { + b.WriteString(` + + +`) + } + + // Footer + b.WriteString(` + + + +
+

📊 Weekly Telemetry Report

+

ProxmoxVE Helper Scripts

+
+ + + + + +
+Calendar Week
+Week `) + b.WriteString(fmt.Sprintf("%d, %d", data.CalendarWeek, data.Year)) + b.WriteString(` +		 +Period
+`) + b.WriteString(fmt.Sprintf("%s – %s", data.StartDate.Format("Jan 02"), data.EndDate.Format("Jan 02, 2006"))) + b.WriteString(` +
+
+ + + + + + + +
+
+
`) + b.WriteString(fmt.Sprintf("%d", data.TotalInstalls)) + b.WriteString(`
+
Total
+
+		 +
+
`) + b.WriteString(fmt.Sprintf("%d", data.SuccessCount)) + b.WriteString(`
+
Successful
+
+		 +
+
`) + b.WriteString(fmt.Sprintf("%d", data.FailedCount)) + b.WriteString(`
+
Failed
+
+		 +
+
`) + b.WriteString(fmt.Sprintf("%.1f%%", data.SuccessRate)) + b.WriteString(`
+
Success Rate
+
+
+
+ + + + + +
+vs. Previous Week
+`) + b.WriteString(installIcon) + b.WriteString(fmt.Sprintf(" %+d installations (%.1f%%)", data.ComparedToPrev.InstallsChange, data.ComparedToPrev.InstallsPercent)) + b.WriteString(` +		 +Failure Rate Change
+`) + b.WriteString(failIcon) + b.WriteString(fmt.Sprintf(" %+.1f percentage points", data.ComparedToPrev.FailRateChange)) + b.WriteString(` +
+
+

🏆 Top 5 Installed Scripts

+ +`) + if len(data.TopApps) > 0 { + for i, app := range data.TopApps { + bgColor := "#ffffff" + if i%2 == 0 { + bgColor = "#f8fafc" + } + b.WriteString(fmt.Sprintf(` + + +`, bgColor, i+1, app.Name, app.Total)) + } + } else { + b.WriteString(``) + } + b.WriteString(`
+%d +%s +%d installs
No data available
+
+

⚠️ Top 5 Scripts with Highest Failure Rates

+ +`) + if len(data.TopFailedApps) > 0 { + for i, app := range data.TopFailedApps { + bgColor := "#ffffff" + if i%2 == 0 { + bgColor = "#fef2f2" + } + rateColor := "#dc2626" + if app.FailureRate < 20 { + rateColor = "#ea580c" + } + if app.FailureRate < 10 { + rateColor = "#ca8a04" + } + b.WriteString(fmt.Sprintf(` + + + +`, bgColor, app.Name, app.Failed, app.Total, rateColor, app.FailureRate)) + } + } else { + b.WriteString(``) + } + b.WriteString(`
+%s +%d / %d failed +%.1f%% +
🎉 No failures this week!
+
+

📦 Distribution by Type

+ + +`) + for t, count := range data.TypeDistribution { + percent := float64(count) / float64(data.TotalInstalls) * 100 + b.WriteString(fmt.Sprintf(``, count, strings.ToUpper(t), percent)) + } + b.WriteString(` +
+
+
%d
+
%s (%.1f%%)
+
+
+
+

🐧 Top Operating Systems

+ +`) + // Sort OS by count + type osEntry struct { + name string + count int + } + var osList []osEntry + for name, count := range data.OsDistribution { + osList = append(osList, osEntry{name, count}) + } + for i := 0; i < len(osList); i++ { + for j := i + 1; j < len(osList); j++ { + if osList[j].count > osList[i].count { + osList[i], osList[j] = osList[j], osList[i] + } + } + } + for i, os := range osList { + if i >= 5 { + break + } + percent := float64(os.count) / float64(data.TotalInstalls) * 100 + barWidth := int(percent * 2) // Scale for visual + if barWidth > 100 { + barWidth = 100 + } + b.WriteString(fmt.Sprintf(` + + + +`, os.name, barWidth, os.count, percent)) + } + b.WriteString(`
%s +
+
+
+		%d (%.1f%%)
+
+

+Generated `) + b.WriteString(time.Now().Format("Jan 02, 2006 at 15:04 MST")) + b.WriteString(`
+ProxmoxVE Helper Scripts — +This is an automated report from the telemetry service. +

+
+
+ +`) + + return b.String() +} + +// generateWeeklyReportEmail creates the plain text email body (kept for compatibility) +func (a *Alerter) generateWeeklyReportEmail(data *WeeklyReportData) string { + var b strings.Builder + + b.WriteString("ProxmoxVE Helper Scripts - Weekly Telemetry Report\n") + b.WriteString("==================================================\n\n") + + b.WriteString(fmt.Sprintf("Calendar Week: %d, %d\n", data.CalendarWeek, data.Year)) + b.WriteString(fmt.Sprintf("Period: %s - %s\n\n", + data.StartDate.Format("Jan 02, 2006"), + data.EndDate.Format("Jan 02, 2006"))) + + b.WriteString("OVERVIEW\n") + b.WriteString("--------\n") + b.WriteString(fmt.Sprintf("Total Installations: %d\n", data.TotalInstalls)) + b.WriteString(fmt.Sprintf("Successful: %d\n", data.SuccessCount)) + b.WriteString(fmt.Sprintf("Failed: %d\n", data.FailedCount)) + b.WriteString(fmt.Sprintf("Success Rate: %.1f%%\n\n", data.SuccessRate)) + + if data.ComparedToPrev.InstallsChange != 0 || data.ComparedToPrev.FailRateChange != 0 { + b.WriteString("vs. Previous Week:\n") + b.WriteString(fmt.Sprintf(" Installations: %+d (%.1f%%)\n", data.ComparedToPrev.InstallsChange, data.ComparedToPrev.InstallsPercent)) + b.WriteString(fmt.Sprintf(" Failure Rate: %+.1f pp\n\n", data.ComparedToPrev.FailRateChange)) + } + + b.WriteString("TOP 5 INSTALLED SCRIPTS\n") + b.WriteString("-----------------------\n") + for i, app := range data.TopApps { + if i >= 5 { + break + } + b.WriteString(fmt.Sprintf("%d. %-25s %5d installs\n", i+1, app.Name, app.Total)) + } + b.WriteString("\n") + + b.WriteString("TOP 5 FAILED SCRIPTS\n") + b.WriteString("--------------------\n") + if len(data.TopFailedApps) > 0 { + for i, app := range data.TopFailedApps { + if i >= 5 { + break + } + b.WriteString(fmt.Sprintf("%d. %-20s %3d/%3d failed (%.1f%%)\n", + i+1, app.Name, app.Failed, app.Total, app.FailureRate)) + } + } else { + b.WriteString("No failures this week!\n") + } + b.WriteString("\n") + + b.WriteString("---\n") + b.WriteString(fmt.Sprintf("Generated: %s\n", time.Now().Format("Jan 02, 2006 15:04 MST"))) + b.WriteString("This is an automated report from the telemetry service.\n") + + return b.String() +} + +// TestWeeklyReport sends a test weekly report email +func (a *Alerter) TestWeeklyReport() error { + return a.SendWeeklyReport() } \ No newline at end of file diff --git a/misc/data/cleanup.go b/misc/data/cleanup.go new file mode 100644 index 000000000..c072bef3d --- /dev/null +++ b/misc/data/cleanup.go @@ -0,0 +1,173 @@ +package main + +import ( + "context" + "encoding/json" + "fmt" + "log" + "net/http" + "net/url" + "time" +) + +// CleanupConfig holds configuration for the cleanup job +type CleanupConfig struct { + Enabled bool + CheckInterval time.Duration // How often to run cleanup + StuckAfterHours int // Consider "installing" as stuck after X hours +} + +// Cleaner handles cleanup of stuck installations +type Cleaner struct { + cfg CleanupConfig + pb *PBClient +} + +// NewCleaner creates a new cleaner instance +func NewCleaner(cfg CleanupConfig, pb *PBClient) *Cleaner { + return &Cleaner{ + cfg: cfg, + pb: pb, + } +} + +// Start begins the cleanup loop +func (c *Cleaner) Start() { + if !c.cfg.Enabled { + log.Println("INFO: cleanup job disabled") + return + } + + go c.cleanupLoop() + log.Printf("INFO: cleanup job started (interval: %v, stuck after: %d hours)", c.cfg.CheckInterval, c.cfg.StuckAfterHours) +} + +func (c *Cleaner) cleanupLoop() { + // Run immediately on start + c.runCleanup() + + ticker := time.NewTicker(c.cfg.CheckInterval) + defer ticker.Stop() + + for range ticker.C { + c.runCleanup() + } +} + +// runCleanup finds and updates stuck installations +func (c *Cleaner) runCleanup() { + ctx, cancel := context.WithTimeout(context.Background(), 60*time.Second) + defer cancel() + + // Find stuck records + stuckRecords, err := c.findStuckInstallations(ctx) + if err != nil { + log.Printf("WARN: cleanup - failed to find stuck installations: %v", err) + return + } + + if len(stuckRecords) == 0 { + log.Printf("INFO: cleanup - no stuck installations found") + return + } + + log.Printf("INFO: cleanup - found %d stuck installations", len(stuckRecords)) + + // Update each record + updated := 0 + for _, record := range stuckRecords { + if err := c.markAsUnknown(ctx, record.ID); err != nil { + log.Printf("WARN: cleanup - failed to update record %s: %v", record.ID, err) + continue + } + updated++ + } + + log.Printf("INFO: cleanup - updated %d stuck installations to 'unknown'", updated) +} + +// StuckRecord represents a minimal record for cleanup +type StuckRecord struct { + ID string `json:"id"` + NSAPP string `json:"nsapp"` + Created string `json:"created"` +} + +// findStuckInstallations finds records that are stuck in "installing" status +func (c *Cleaner) findStuckInstallations(ctx context.Context) ([]StuckRecord, error) { + if err := c.pb.ensureAuth(ctx); err != nil { + return nil, err + } + + // Calculate cutoff time + cutoff := time.Now().Add(-time.Duration(c.cfg.StuckAfterHours) * time.Hour) + cutoffStr := cutoff.Format("2006-01-02 15:04:05") + + // Build filter: status='installing' AND created < cutoff + filter := url.QueryEscape(fmt.Sprintf("status='installing' && created<'%s'", cutoffStr)) + + req, err := http.NewRequestWithContext(ctx, http.MethodGet, + fmt.Sprintf("%s/api/collections/%s/records?filter=%s&perPage=100", + c.pb.baseURL, c.pb.devColl, filter), + nil, + ) + if err != nil { + return nil, err + } + req.Header.Set("Authorization", "Bearer "+c.pb.token) + + resp, err := c.pb.http.Do(req) + if err != nil { + return nil, err + } + defer resp.Body.Close() + + var result struct { + Items []StuckRecord `json:"items"` + } + if err := json.NewDecoder(resp.Body).Decode(&result); err != nil { + return nil, err + } + + return result.Items, nil +} + +// markAsUnknown updates a record's status to "unknown" +func (c *Cleaner) markAsUnknown(ctx context.Context, recordID string) error { + update := TelemetryStatusUpdate{ + Status: "unknown", + Error: "Installation timed out - no completion status received", + } + return c.pb.UpdateTelemetryStatus(ctx, recordID, update) +} + +// RunNow triggers an immediate cleanup run (for testing/manual trigger) +func (c *Cleaner) RunNow() (int, error) { + ctx, cancel := context.WithTimeout(context.Background(), 60*time.Second) + defer cancel() + + stuckRecords, err := c.findStuckInstallations(ctx) + if err != nil { + return 0, fmt.Errorf("failed to find stuck installations: %w", err) + } + + updated := 0 + for _, record := range stuckRecords { + if err := c.markAsUnknown(ctx, record.ID); err != nil { + log.Printf("WARN: cleanup - failed to update record %s: %v", record.ID, err) + continue + } + updated++ + } + + return updated, nil +} + +// GetStuckCount returns the current number of stuck installations +func (c *Cleaner) GetStuckCount(ctx context.Context) (int, error) { + records, err := c.findStuckInstallations(ctx) + if err != nil { + return 0, err + } + return len(records), nil +} diff --git a/misc/data/dashboard.go b/misc/data/dashboard.go index 6b28625e4..ab899be94 100644 --- a/misc/data/dashboard.go +++ b/misc/data/dashboard.go @@ -111,12 +111,17 @@ func (p *PBClient) FetchDashboardData(ctx context.Context, days int) (*Dashboard data := &DashboardData{} - // Calculate date filter - since := time.Now().AddDate(0, 0, -days).Format("2006-01-02 00:00:00") - filter := url.QueryEscape(fmt.Sprintf("created >= '%s'", since)) + // Calculate date filter (days=0 means all entries) + var filter string + if days > 0 { + since := time.Now().AddDate(0, 0, -days).Format("2006-01-02 00:00:00") + filter = url.QueryEscape(fmt.Sprintf("created >= '%s'", since)) + } else { + filter = "" // No filter = all entries + } // Fetch all records for the period - records, err := p.fetchRecords(ctx, filter, 500) + records, err := p.fetchRecords(ctx, filter) if err != nil { return nil, err } @@ -292,17 +297,22 @@ type TelemetryRecord struct { Created string `json:"created"` } -func (p *PBClient) fetchRecords(ctx context.Context, filter string, limit int) ([]TelemetryRecord, error) { +func (p *PBClient) fetchRecords(ctx context.Context, filter string) ([]TelemetryRecord, error) { var allRecords []TelemetryRecord page := 1 - perPage := 100 + perPage := 500 for { - req, err := http.NewRequestWithContext(ctx, http.MethodGet, - fmt.Sprintf("%s/api/collections/%s/records?filter=%s&sort=-created&page=%d&perPage=%d", - p.baseURL, p.targetColl, filter, page, perPage), - nil, - ) + var url string + if filter != "" { + url = fmt.Sprintf("%s/api/collections/%s/records?filter=%s&sort=-created&page=%d&perPage=%d", + p.baseURL, p.devColl, filter, page, perPage) + } else { + url = fmt.Sprintf("%s/api/collections/%s/records?sort=-created&page=%d&perPage=%d", + p.baseURL, p.devColl, page, perPage) + } + + req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil) if err != nil { return nil, err } @@ -325,7 +335,7 @@ func (p *PBClient) fetchRecords(ctx context.Context, filter string, limit int) ( allRecords = append(allRecords, result.Items...) - if len(allRecords) >= limit || len(allRecords) >= result.TotalItems { + if len(allRecords) >= result.TotalItems { break } page++ @@ -746,6 +756,36 @@ func DashboardHTML() string { color: #fff; } + .quickfilter { + display: flex; + gap: 4px; + background: var(--bg-tertiary); + padding: 4px; + border-radius: 8px; + border: 1px solid var(--border-color); + } + + .filter-btn { + background: transparent; + border: none; + color: var(--text-secondary); + padding: 6px 12px; + border-radius: 6px; + font-size: 13px; + font-weight: 500; + transition: all 0.2s; + } + + .filter-btn:hover { + background: var(--bg-secondary); + color: var(--text-primary); + } + + .filter-btn.active { + background: var(--accent-blue); + color: #fff; + } + .stats-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(200px, 1fr)); @@ -951,6 +991,98 @@ func DashboardHTML() string { background: var(--bg-secondary); } + .admin-btn { + background: var(--bg-tertiary); + border: 1px solid var(--border-color); + color: var(--text-primary); + padding: 4px 8px; + border-radius: 4px; + font-size: 12px; + cursor: pointer; + margin-left: 8px; + } + + .admin-btn:hover { + background: var(--accent-blue); + color: #fff; + border-color: var(--accent-blue); + } + + .admin-btn:disabled { + opacity: 0.5; + cursor: not-allowed; + } + + .footer-btn { + background: transparent; + border: none; + color: var(--accent-blue); + cursor: pointer; + font-size: 12px; + padding: 0; + margin-right: 8px; + } + + .footer-btn:hover { + text-decoration: underline; + } + + .health-modal { + max-width: 400px; + } + + .health-status { + display: flex; + align-items: center; + gap: 12px; + padding: 16px; + border-radius: 8px; + margin-bottom: 12px; + } + + .health-status.ok { + background: rgba(63, 185, 80, 0.1); + border: 1px solid var(--accent-green); + } + + .health-status.error { + background: rgba(248, 81, 73, 0.1); + border: 1px solid var(--accent-red); + } + + .health-status .icon { + font-size: 32px; + } + + .health-status .details { + flex: 1; + } + + .health-status .title { + font-weight: 600; + font-size: 16px; + } + + .health-status .subtitle { + font-size: 12px; + color: var(--text-secondary); + margin-top: 4px; + } + + .health-info { + font-size: 12px; + color: var(--text-secondary); + padding: 12px; + background: var(--bg-tertiary); + border-radius: 6px; + } + + .health-info div { + display: flex; + justify-content: space-between; + padding: 4px 0; + } + .pve-version-card { background: var(--bg-secondary); border: 1px solid var(--border-color); @@ -1104,6 +1236,171 @@ func DashboardHTML() string { color: var(--text-secondary); font-size: 14px; } + + /* Detail Modal Styles */ + .modal-overlay { + position: fixed; + top: 0; + left: 0; + right: 0; + bottom: 0; + background: rgba(0, 0, 0, 0.7); + display: flex; + justify-content: center; + align-items: center; + z-index: 1000; + opacity: 0; + visibility: hidden; + transition: opacity 0.2s, visibility 0.2s; + } + + .modal-overlay.active { + opacity: 1; + visibility: visible; + } + + .modal-content { + background: var(--bg-secondary); + border: 1px solid var(--border-color); + border-radius: 12px; + width: 90%; + max-width: 700px; + max-height: 90vh; + overflow-y: auto; + transform: scale(0.9); + transition: transform 0.2s; + } + + .modal-overlay.active .modal-content { + transform: scale(1); + } + + .modal-header { + display: flex; + justify-content: space-between; + align-items: center; + padding: 20px 24px; + border-bottom: 1px solid var(--border-color); + position: sticky; + top: 0; + background: var(--bg-secondary); + z-index: 10; + } + + .modal-header h2 { + font-size: 20px; + font-weight: 600; + display: flex; + align-items: center; + gap: 10px; + } + + .modal-close { + background: none; + border: none; + color: var(--text-secondary); + font-size: 24px; + cursor: pointer; + padding: 4px 8px; + border-radius: 4px; + } + + .modal-close:hover { + background: var(--bg-tertiary); + color: var(--text-primary); + } + + .modal-body { + padding: 24px; + } + + .detail-section { + margin-bottom: 24px; + } + + .detail-section:last-child { + margin-bottom: 0; + } + + .detail-section-header { + display: flex; + align-items: center; + gap: 8px; + font-size: 12px; + font-weight: 600; + color: var(--text-secondary); + text-transform: uppercase; + letter-spacing: 0.5px; + margin-bottom: 12px; + padding-bottom: 8px; + border-bottom: 1px solid var(--border-color); + } + + .detail-section-header svg { + opacity: 0.7; + } + + .detail-grid { + display: grid; + grid-template-columns: repeat(auto-fit, minmax(200px, 1fr)); + gap: 16px; + } + + .detail-item { + background: var(--bg-tertiary); + border-radius: 8px; + padding: 12px 16px; + } + + .detail-item .label { + font-size: 11px; + color: var(--text-secondary); + text-transform: uppercase; + letter-spacing: 0.3px; + margin-bottom: 4px; + } + + .detail-item .value { + font-size: 15px; + font-weight: 500; + word-break: break-word; + } + + .detail-item .value.mono { + font-family: 'SF Mono', 'Consolas', monospace; + font-size: 13px; + } + + .detail-item.full-width { + grid-column: 1 / -1; + } + + .detail-item .value.status-success { color: var(--accent-green); } + .detail-item .value.status-failed { color: var(--accent-red); } + .detail-item .value.status-installing { color: var(--accent-yellow); } + + .error-box { + background: rgba(248, 81, 73, 0.1); + border: 1px solid rgba(248, 81, 73, 0.3); + border-radius: 8px; + padding: 16px; + font-family: 'SF Mono', 'Consolas', monospace; + font-size: 13px; + color: var(--accent-red); + white-space: pre-wrap; + word-break: break-word; + max-height: 200px; + overflow-y: auto; + } + + tr.clickable-row { + cursor: pointer; + transition: background 0.15s; + } + + tr.clickable-row:hover { + background: rgba(88, 166, 255, 0.1) !important; + } @@ -1116,13 +1413,13 @@ func DashboardHTML() string { Telemetry Dashboard
- +
+ + + + + +
- Health Check • - Metrics • + API
+ + + + + +