1 Commits

Author SHA1 Message Date
lerko e36d9a7a26 chore: add TUI screenshots via VHS
CI / test (pull_request) Successful in 2m47s
CI / lint (pull_request) Successful in 1m21s
CI / vulncheck (pull_request) Successful in 57s
Monitors dashboard, detail panel, alerts, and logs views.
Generated with charmbracelet/vhs using real monitor data.
2026-05-27 19:03:03 -04:00
31 changed files with 217 additions and 1390 deletions
+4
View File
@@ -37,3 +37,7 @@ tmp
*.local.json *.local.json
*.local.md *.local.md
# VHS — seed has personal URLs, tape is local workflow
vhs/seed.yaml
vhs/demo.tape
+7 -9
View File
@@ -61,15 +61,13 @@ func (m *mockStore) GetActiveMaintenanceWindows() ([]models.MaintenanceWindow, e
func (m *mockStore) GetAllMaintenanceWindows(int) ([]models.MaintenanceWindow, error) { func (m *mockStore) GetAllMaintenanceWindows(int) ([]models.MaintenanceWindow, error) {
return nil, nil return nil, nil
} }
func (m *mockStore) AddMaintenanceWindow(models.MaintenanceWindow) error { return nil } func (m *mockStore) AddMaintenanceWindow(models.MaintenanceWindow) error { return nil }
func (m *mockStore) EndMaintenanceWindow(int) error { return nil } func (m *mockStore) EndMaintenanceWindow(int) error { return nil }
func (m *mockStore) DeleteMaintenanceWindow(int) error { return nil } func (m *mockStore) DeleteMaintenanceWindow(int) error { return nil }
func (m *mockStore) IsMonitorInMaintenance(int) (bool, error) { return false, nil } func (m *mockStore) IsMonitorInMaintenance(int) (bool, error) { return false, nil }
func (m *mockStore) GetPreference(string) (string, error) { return "", nil } func (m *mockStore) GetPreference(string) (string, error) { return "", nil }
func (m *mockStore) SetPreference(string, string) error { return nil } func (m *mockStore) SetPreference(string, string) error { return nil }
func (m *mockStore) SaveStateChange(int, string, string, string) error { return nil } func (m *mockStore) Close() error { return nil }
func (m *mockStore) GetStateChanges(int, int) ([]models.StateChange, error) { return nil, nil }
func (m *mockStore) Close() error { return nil }
// --- Cluster Start Tests --- // --- Cluster Start Tests ---
+6 -8
View File
@@ -127,10 +127,9 @@ func probeFetchAssignments(ctx context.Context, client *http.Client, cfg ProbeCo
} }
type probeResultItem struct { type probeResultItem struct {
SiteID int `json:"site_id"` SiteID int `json:"site_id"`
LatencyNs int64 `json:"latency_ns"` LatencyNs int64 `json:"latency_ns"`
IsUp bool `json:"is_up"` IsUp bool `json:"is_up"`
ErrorReason string `json:"error_reason,omitempty"`
} }
func probeExecuteChecks(ctx context.Context, sites []models.Site, strict, insecure *http.Client, allowPrivate bool) []probeResultItem { func probeExecuteChecks(ctx context.Context, sites []models.Site, strict, insecure *http.Client, allowPrivate bool) []probeResultItem {
@@ -155,10 +154,9 @@ loop:
cr := monitor.RunCheck(s, strict, insecure, false, allowPrivate) cr := monitor.RunCheck(s, strict, insecure, false, allowPrivate)
mu.Lock() mu.Lock()
results = append(results, probeResultItem{ results = append(results, probeResultItem{
SiteID: s.ID, SiteID: s.ID,
LatencyNs: cr.LatencyNs, LatencyNs: cr.LatencyNs,
IsUp: cr.Status == "UP", IsUp: cr.Status == "UP",
ErrorReason: cr.ErrorReason,
}) })
mu.Unlock() mu.Unlock()
}(site) }(site)
+9 -12
View File
@@ -2,14 +2,13 @@ package metrics
import ( import (
"context" "context"
"gitea.lerkolabs.com/lerko/uptop/internal/models"
"gitea.lerkolabs.com/lerko/uptop/internal/monitor"
"net/http" "net/http"
"net/http/httptest" "net/http/httptest"
"strings" "strings"
"testing" "testing"
"time" "time"
"gitea.lerkolabs.com/lerko/uptop/internal/models"
"gitea.lerkolabs.com/lerko/uptop/internal/monitor"
) )
type mockStore struct { type mockStore struct {
@@ -59,15 +58,13 @@ func (m *mockStore) GetActiveMaintenanceWindows() ([]models.MaintenanceWindow, e
func (m *mockStore) GetAllMaintenanceWindows(int) ([]models.MaintenanceWindow, error) { func (m *mockStore) GetAllMaintenanceWindows(int) ([]models.MaintenanceWindow, error) {
return nil, nil return nil, nil
} }
func (m *mockStore) AddMaintenanceWindow(models.MaintenanceWindow) error { return nil } func (m *mockStore) AddMaintenanceWindow(models.MaintenanceWindow) error { return nil }
func (m *mockStore) EndMaintenanceWindow(int) error { return nil } func (m *mockStore) EndMaintenanceWindow(int) error { return nil }
func (m *mockStore) DeleteMaintenanceWindow(int) error { return nil } func (m *mockStore) DeleteMaintenanceWindow(int) error { return nil }
func (m *mockStore) IsMonitorInMaintenance(int) (bool, error) { return false, nil } func (m *mockStore) IsMonitorInMaintenance(int) (bool, error) { return false, nil }
func (m *mockStore) GetPreference(string) (string, error) { return "", nil } func (m *mockStore) GetPreference(string) (string, error) { return "", nil }
func (m *mockStore) SetPreference(string, string) error { return nil } func (m *mockStore) SetPreference(string, string) error { return nil }
func (m *mockStore) SaveStateChange(int, string, string, string) error { return nil } func (m *mockStore) Close() error { return nil }
func (m *mockStore) GetStateChanges(int, int) ([]models.StateChange, error) { return nil, nil }
func (m *mockStore) Close() error { return nil }
func TestMetricsHandler(t *testing.T) { func TestMetricsHandler(t *testing.T) {
ms := &mockStore{ ms := &mockStore{
+8 -20
View File
@@ -27,26 +27,14 @@ type Site struct {
Paused bool Paused bool
Regions string Regions string
FailureCount int FailureCount int
Status string Status string
StatusCode int StatusCode int
Latency time.Duration Latency time.Duration
CertExpiry time.Time CertExpiry time.Time
HasSSL bool HasSSL bool
LastCheck time.Time LastCheck time.Time
SentSSLWarning bool SentSSLWarning bool
LastError string
StatusChangedAt time.Time
LastSuccessAt time.Time
}
type StateChange struct {
ID int
SiteID int
FromStatus string
ToStatus string
ErrorReason string
ChangedAt time.Time
} }
type AlertConfig struct { type AlertConfig struct {
+4 -5
View File
@@ -11,11 +11,10 @@ const (
) )
type NodeResult struct { type NodeResult struct {
NodeID string NodeID string
IsUp bool IsUp bool
LatencyNs int64 LatencyNs int64
CheckedAt time.Time CheckedAt time.Time
ErrorReason string
} }
func AggregateStatus(results []NodeResult, strategy AggregationStrategy) (isUp bool, avgLatencyNs int64) { func AggregateStatus(results []NodeResult, strategy AggregationStrategy) (isUp bool, avgLatencyNs int64) {
+15 -34
View File
@@ -2,7 +2,6 @@ package monitor
import ( import (
"context" "context"
"fmt"
"net" "net"
"net/http" "net/http"
"strconv" "strconv"
@@ -16,13 +15,12 @@ import (
) )
type CheckResult struct { type CheckResult struct {
SiteID int SiteID int
Status string // "UP", "DOWN", "SSL EXP" Status string // "UP", "DOWN", "SSL EXP"
StatusCode int StatusCode int
LatencyNs int64 LatencyNs int64
HasSSL bool HasSSL bool
CertExpiry time.Time CertExpiry time.Time
ErrorReason string
} }
func RunCheck(site models.Site, strict, insecure *http.Client, globalInsecure bool, allowPrivate ...bool) CheckResult { func RunCheck(site models.Site, strict, insecure *http.Client, globalInsecure bool, allowPrivate ...bool) CheckResult {
@@ -37,7 +35,7 @@ func RunCheck(site models.Site, strict, insecure *http.Client, globalInsecure bo
if ips, err := net.LookupIP(host); err == nil { if ips, err := net.LookupIP(host); err == nil {
for _, ip := range ips { for _, ip := range ips {
if isPrivateIP(ip) { if isPrivateIP(ip) {
return CheckResult{SiteID: site.ID, Status: "DOWN", ErrorReason: "target resolves to private IP"} return CheckResult{SiteID: site.ID, Status: "DOWN"}
} }
} }
} }
@@ -54,7 +52,7 @@ func RunCheck(site models.Site, strict, insecure *http.Client, globalInsecure bo
case "dns": case "dns":
return runDNSCheck(site) return runDNSCheck(site)
default: default:
return CheckResult{SiteID: site.ID, Status: "DOWN", ErrorReason: "unsupported monitor type: " + site.Type} return CheckResult{SiteID: site.ID, Status: "DOWN"}
} }
} }
@@ -70,7 +68,7 @@ func runHTTPCheck(site models.Site, strict, insecure *http.Client, globalInsecur
req, err := http.NewRequestWithContext(ctx, method, site.URL, nil) req, err := http.NewRequestWithContext(ctx, method, site.URL, nil)
if err != nil { if err != nil {
return CheckResult{SiteID: site.ID, Status: "DOWN", ErrorReason: "invalid request: " + err.Error()} return CheckResult{SiteID: site.ID, Status: "DOWN"}
} }
client := strict client := strict
@@ -90,7 +88,6 @@ func runHTTPCheck(site models.Site, strict, insecure *http.Client, globalInsecur
if err != nil { if err != nil {
result.Status = "DOWN" result.Status = "DOWN"
result.ErrorReason = truncateError(err.Error(), 256)
return result return result
} }
defer resp.Body.Close() defer resp.Body.Close()
@@ -98,11 +95,6 @@ func runHTTPCheck(site models.Site, strict, insecure *http.Client, globalInsecur
result.StatusCode = resp.StatusCode result.StatusCode = resp.StatusCode
if !isCodeAccepted(resp.StatusCode, site.AcceptedCodes) { if !isCodeAccepted(resp.StatusCode, site.AcceptedCodes) {
result.Status = "DOWN" result.Status = "DOWN"
expected := site.AcceptedCodes
if expected == "" {
expected = "200-299"
}
result.ErrorReason = fmt.Sprintf("HTTP %d (expected %s)", resp.StatusCode, expected)
} }
if site.CheckSSL && resp.TLS != nil && len(resp.TLS.PeerCertificates) > 0 { if site.CheckSSL && resp.TLS != nil && len(resp.TLS.PeerCertificates) > 0 {
@@ -111,7 +103,6 @@ func runHTTPCheck(site models.Site, strict, insecure *http.Client, globalInsecur
result.CertExpiry = cert.NotAfter result.CertExpiry = cert.NotAfter
if time.Now().After(cert.NotAfter) { if time.Now().After(cert.NotAfter) {
result.Status = "SSL EXP" result.Status = "SSL EXP"
result.ErrorReason = "SSL certificate expired"
} }
} }
@@ -126,7 +117,7 @@ func runPingCheck(site models.Site) CheckResult {
pinger, err := probing.NewPinger(host) pinger, err := probing.NewPinger(host)
if err != nil { if err != nil {
return CheckResult{SiteID: site.ID, Status: "DOWN", ErrorReason: "ping setup: " + err.Error()} return CheckResult{SiteID: site.ID, Status: "DOWN"}
} }
pinger.Count = 1 pinger.Count = 1
pinger.Timeout = siteTimeout(site) pinger.Timeout = siteTimeout(site)
@@ -136,11 +127,8 @@ func runPingCheck(site models.Site) CheckResult {
err = pinger.Run() err = pinger.Run()
latency := time.Since(start) latency := time.Since(start)
if err != nil { if err != nil || pinger.Statistics().PacketsRecv == 0 {
return CheckResult{SiteID: site.ID, Status: "DOWN", LatencyNs: latency.Nanoseconds(), ErrorReason: "ping failed: " + err.Error()} return CheckResult{SiteID: site.ID, Status: "DOWN", LatencyNs: latency.Nanoseconds()}
}
if pinger.Statistics().PacketsRecv == 0 {
return CheckResult{SiteID: site.ID, Status: "DOWN", LatencyNs: latency.Nanoseconds(), ErrorReason: "no ICMP response"}
} }
stats := pinger.Statistics() stats := pinger.Statistics()
@@ -160,7 +148,7 @@ func runPortCheck(site models.Site) CheckResult {
latency := time.Since(start) latency := time.Since(start)
if err != nil { if err != nil {
return CheckResult{SiteID: site.ID, Status: "DOWN", LatencyNs: latency.Nanoseconds(), ErrorReason: truncateError(err.Error(), 256)} return CheckResult{SiteID: site.ID, Status: "DOWN", LatencyNs: latency.Nanoseconds()}
} }
_ = conn.Close() _ = conn.Close()
return CheckResult{SiteID: site.ID, Status: "UP", LatencyNs: latency.Nanoseconds()} return CheckResult{SiteID: site.ID, Status: "UP", LatencyNs: latency.Nanoseconds()}
@@ -211,10 +199,10 @@ func runDNSCheck(site models.Site) CheckResult {
latency := time.Since(start) latency := time.Since(start)
if err != nil { if err != nil {
return CheckResult{SiteID: site.ID, Status: "DOWN", LatencyNs: latency.Nanoseconds(), ErrorReason: "DNS query failed: " + err.Error()} return CheckResult{SiteID: site.ID, Status: "DOWN", LatencyNs: latency.Nanoseconds()}
} }
if r.Rcode != dns.RcodeSuccess { if r.Rcode != dns.RcodeSuccess {
return CheckResult{SiteID: site.ID, Status: "DOWN", StatusCode: r.Rcode, LatencyNs: latency.Nanoseconds(), ErrorReason: "DNS RCODE: " + dns.RcodeToString[r.Rcode]} return CheckResult{SiteID: site.ID, Status: "DOWN", StatusCode: r.Rcode, LatencyNs: latency.Nanoseconds()}
} }
return CheckResult{SiteID: site.ID, Status: "UP", LatencyNs: latency.Nanoseconds()} return CheckResult{SiteID: site.ID, Status: "UP", LatencyNs: latency.Nanoseconds()}
} }
@@ -247,10 +235,3 @@ func isCodeAccepted(code int, accepted string) bool {
} }
return false return false
} }
func truncateError(s string, max int) string {
if len(s) <= max {
return s
}
return s[:max-3] + "..."
}
+25 -181
View File
@@ -19,18 +19,10 @@ import (
const ( const (
maxLogEntries = 100 maxLogEntries = 100
pollInterval = 5 * time.Second pollInterval = 5 * time.Second
pushGracePeriod = 5 * time.Second
minCheckInterval = 5 minCheckInterval = 5
minPushGrace = 60 * time.Second
) )
type AlertHealth struct {
LastSendAt time.Time
LastSendOK bool
LastError string
SendCount int
FailCount int
}
type Engine struct { type Engine struct {
mu sync.RWMutex mu sync.RWMutex
liveState map[int]models.Site liveState map[int]models.Site
@@ -50,9 +42,6 @@ type Engine struct {
probeResults map[int]map[string]NodeResult probeResults map[int]map[string]NodeResult
aggStrategy AggregationStrategy aggStrategy AggregationStrategy
alertHealthMu sync.RWMutex
alertHealth map[int]AlertHealth
db store.Store db store.Store
insecureSkipVerify bool insecureSkipVerify bool
allowPrivateTargets bool allowPrivateTargets bool
@@ -75,7 +64,6 @@ func newEngine(s store.Store, allowPrivateTargets bool) *Engine {
histories: make(map[int]*SiteHistory), histories: make(map[int]*SiteHistory),
tokenIndex: make(map[string]int), tokenIndex: make(map[string]int),
probeResults: make(map[int]map[string]NodeResult), probeResults: make(map[int]map[string]NodeResult),
alertHealth: make(map[int]AlertHealth),
aggStrategy: AggAnyDown, aggStrategy: AggAnyDown,
isActive: true, isActive: true,
allowPrivateTargets: allowPrivateTargets, allowPrivateTargets: allowPrivateTargets,
@@ -108,19 +96,6 @@ func sanitizeLog(s string) string {
return s return s
} }
func fmtDurationShort(d time.Duration) string {
if d < time.Minute {
return fmt.Sprintf("%ds", int(d.Seconds()))
}
if d < time.Hour {
return fmt.Sprintf("%dm", int(d.Minutes()))
}
if d < 24*time.Hour {
return fmt.Sprintf("%dh %dm", int(d.Hours()), int(d.Minutes())%60)
}
return fmt.Sprintf("%dd %dh", int(d.Hours())/24, int(d.Hours())%24)
}
func (e *Engine) AddLog(msg string) { func (e *Engine) AddLog(msg string) {
e.logMu.Lock() e.logMu.Lock()
defer e.logMu.Unlock() defer e.logMu.Unlock()
@@ -211,38 +186,17 @@ func (e *Engine) RecordHeartbeat(token string) bool {
return false return false
} }
prevStatus := site.Status
site.LastCheck = time.Now() site.LastCheck = time.Now()
wasDown := site.Status == "DOWN"
site.Status = "UP" site.Status = "UP"
site.FailureCount = 0 site.FailureCount = 0
site.Latency = 0 site.Latency = 0
site.LastError = ""
site.LastSuccessAt = time.Now()
if prevStatus != "UP" {
site.StatusChangedAt = time.Now()
}
e.liveState[targetID] = site e.liveState[targetID] = site
switch prevStatus { if wasDown {
case "PENDING": e.AddLog(fmt.Sprintf("Push Monitor '%s' recovered", site.Name))
e.AddLog(fmt.Sprintf("Push Monitor '%s' received first heartbeat", site.Name)) e.triggerAlert(site.AlertID, "✅ RECOVERY", fmt.Sprintf("Push Monitor '%s' is receiving heartbeats.", site.Name))
case "LATE":
e.AddLog(fmt.Sprintf("Push Monitor '%s' heartbeat arrived (was late)", site.Name))
case "DOWN":
downDur := ""
if !site.StatusChangedAt.IsZero() {
downDur = fmt.Sprintf(" (was down %s)", fmtDurationShort(time.Since(site.StatusChangedAt)))
}
e.AddLog(fmt.Sprintf("Push Monitor '%s' recovered%s", site.Name, downDur))
go e.triggerAlert(site.AlertID, "✅ RECOVERY", fmt.Sprintf("Push Monitor '%s' is receiving heartbeats.%s", site.Name, downDur))
} }
if prevStatus != "UP" && prevStatus != "PENDING" {
go func() { _ = e.db.SaveStateChange(targetID, prevStatus, "UP", "") }()
}
return true return true
} }
@@ -287,6 +241,9 @@ func (e *Engine) Start(ctx context.Context) {
if !exists { if !exists {
e.mu.Lock() e.mu.Lock()
s.Status = "PENDING" s.Status = "PENDING"
if s.Type == "push" {
s.LastCheck = time.Now()
}
if h, ok := e.GetHistory(s.ID); ok && len(h.Statuses) > 0 { if h, ok := e.GetHistory(s.ID); ok && len(h.Statuses) > 0 {
if h.Statuses[len(h.Statuses)-1] { if h.Statuses[len(h.Statuses)-1] {
s.Status = "UP" s.Status = "UP"
@@ -326,9 +283,6 @@ func (e *Engine) UpdateSiteConfig(site models.Site) {
site.LastCheck = existing.LastCheck site.LastCheck = existing.LastCheck
site.SentSSLWarning = existing.SentSSLWarning site.SentSSLWarning = existing.SentSSLWarning
site.FailureCount = existing.FailureCount site.FailureCount = existing.FailureCount
site.LastError = existing.LastError
site.StatusChangedAt = existing.StatusChangedAt
site.LastSuccessAt = existing.LastSuccessAt
e.liveState[site.ID] = site e.liveState[site.ID] = site
e.addToTokenIndex(site) e.addToTokenIndex(site)
} }
@@ -439,62 +393,33 @@ func (e *Engine) checkByID(id int) {
updatedSite.CertExpiry = result.CertExpiry updatedSite.CertExpiry = result.CertExpiry
updatedSite.Latency = time.Duration(result.LatencyNs) updatedSite.Latency = time.Duration(result.LatencyNs)
updatedSite.LastCheck = time.Now() updatedSite.LastCheck = time.Now()
e.handleStatusChange(updatedSite, result.Status, result.StatusCode, time.Duration(result.LatencyNs), result.ErrorReason) e.handleStatusChange(updatedSite, result.Status, result.StatusCode, time.Duration(result.LatencyNs))
} }
} }
func (e *Engine) checkPush(site models.Site) { func (e *Engine) checkPush(site models.Site) {
if site.Status == "PENDING" { deadline := site.LastCheck.Add(time.Duration(site.Interval) * time.Second).Add(pushGracePeriod)
return if time.Now().After(deadline) {
} e.handleStatusChange(site, "DOWN", 0, 0)
} else if site.Status != "UP" {
interval := time.Duration(site.Interval) * time.Second e.handleStatusChange(site, "UP", 200, 0)
grace := interval / 2
if grace < minPushGrace {
grace = minPushGrace
}
overdue := site.LastCheck.Add(interval)
graceEnd := overdue.Add(grace)
now := time.Now()
if now.After(graceEnd) {
if site.Status != "DOWN" {
e.handleStatusChange(site, "DOWN", 0, 0, "heartbeat missed")
}
} else if now.After(overdue) {
if site.Status != "LATE" {
e.handleStatusChange(site, "LATE", 0, 0, "heartbeat overdue")
}
} }
} }
func (e *Engine) handleStatusChange(site models.Site, rawStatus string, code int, latency time.Duration, errorReason string) { func (e *Engine) handleStatusChange(site models.Site, rawStatus string, code int, latency time.Duration) {
if !e.IsActive() { if !e.IsActive() {
return return
} }
newState := site newState := site
newState.StatusCode = code newState.StatusCode = code
newState.LastError = errorReason
if rawStatus == "UP" {
newState.LastSuccessAt = time.Now()
newState.LastError = ""
} else {
newState.LastSuccessAt = site.LastSuccessAt
}
if site.Status == "UP" && rawStatus != "UP" { if site.Status == "UP" && rawStatus != "UP" {
newState.FailureCount++ newState.FailureCount++
if newState.FailureCount > site.MaxRetries { if newState.FailureCount > site.MaxRetries {
newState.Status = rawStatus newState.Status = rawStatus
newState.FailureCount = site.MaxRetries + 1 newState.FailureCount = site.MaxRetries + 1
if errorReason != "" { e.AddLog(fmt.Sprintf("Monitor '%s' confirmed DOWN", site.Name))
e.AddLog(fmt.Sprintf("Monitor '%s' confirmed DOWN: %s", site.Name, errorReason))
} else {
e.AddLog(fmt.Sprintf("Monitor '%s' confirmed DOWN", site.Name))
}
} else { } else {
e.AddLog(fmt.Sprintf("Monitor '%s' failed check %d/%d", site.Name, newState.FailureCount, site.MaxRetries)) e.AddLog(fmt.Sprintf("Monitor '%s' failed check %d/%d", site.Name, newState.FailureCount, site.MaxRetries))
} }
@@ -506,14 +431,6 @@ func (e *Engine) handleStatusChange(site models.Site, rawStatus string, code int
newState.FailureCount = site.MaxRetries + 1 newState.FailureCount = site.MaxRetries + 1
} }
if newState.Status != site.Status && site.Status != "PENDING" {
newState.StatusChangedAt = time.Now()
} else if site.StatusChangedAt.IsZero() && newState.Status != "PENDING" {
newState.StatusChangedAt = time.Now()
} else {
newState.StatusChangedAt = site.StatusChangedAt
}
inMaint := e.isInMaintenance(site.ID) inMaint := e.isInMaintenance(site.ID)
if site.Type == "http" && site.CheckSSL && site.HasSSL { if site.Type == "http" && site.CheckSSL && site.HasSSL {
@@ -538,24 +455,12 @@ func (e *Engine) handleStatusChange(site models.Site, rawStatus string, code int
e.recordCheck(site.ID, latency, rawStatus == "UP") e.recordCheck(site.ID, latency, rawStatus == "UP")
if newState.Status != site.Status && site.Status != "PENDING" {
go func() { _ = e.db.SaveStateChange(site.ID, site.Status, newState.Status, errorReason) }()
}
isBroken := func(s string) bool { return s == "DOWN" || s == "SSL EXP" } isBroken := func(s string) bool { return s == "DOWN" || s == "SSL EXP" }
if site.Status == "UP" && newState.Status == "LATE" {
e.AddLog(fmt.Sprintf("Monitor '%s' heartbeat overdue", site.Name))
}
if !isBroken(site.Status) && isBroken(newState.Status) && newState.Status != "PENDING" { if !isBroken(site.Status) && isBroken(newState.Status) && newState.Status != "PENDING" {
if inMaint { if inMaint {
e.AddLog(fmt.Sprintf("Monitor '%s' is DOWN (alerts suppressed — maintenance)", site.Name)) e.AddLog(fmt.Sprintf("Monitor '%s' is DOWN (alerts suppressed — maintenance)", site.Name))
} else { } else {
msg := fmt.Sprintf("Monitor '%s' is DOWN (%s)", site.Name, rawStatus) msg := fmt.Sprintf("Monitor '%s' is DOWN (%s)", site.Name, rawStatus)
if errorReason != "" {
msg = fmt.Sprintf("Monitor '%s' is DOWN: %s", site.Name, errorReason)
}
if site.Type == "push" { if site.Type == "push" {
msg = fmt.Sprintf("Push Monitor '%s' missed heartbeat.", site.Name) msg = fmt.Sprintf("Push Monitor '%s' missed heartbeat.", site.Name)
} }
@@ -563,18 +468,12 @@ func (e *Engine) handleStatusChange(site models.Site, rawStatus string, code int
} }
} }
if isBroken(site.Status) && newState.Status == "UP" { if isBroken(site.Status) && newState.Status == "UP" {
downDur := ""
if !site.StatusChangedAt.IsZero() {
downDur = fmt.Sprintf(" (was down %s)", fmtDurationShort(time.Since(site.StatusChangedAt)))
}
e.AddLog(fmt.Sprintf("Monitor '%s' recovered%s", site.Name, downDur))
if !inMaint { if !inMaint {
e.triggerAlert(site.AlertID, "✅ RECOVERY", fmt.Sprintf("Monitor '%s' is UP%s", site.Name, downDur)) e.triggerAlert(site.AlertID, "✅ RECOVERY", fmt.Sprintf("Monitor '%s' is UP", site.Name))
} else {
e.AddLog(fmt.Sprintf("Monitor '%s' recovered (maintenance active, alert suppressed)", site.Name))
} }
} }
if site.Status == "LATE" && newState.Status == "UP" && !isBroken(site.Status) {
e.AddLog(fmt.Sprintf("Monitor '%s' heartbeat arrived (was late)", site.Name))
}
} }
func (e *Engine) triggerAlert(alertID int, title, message string) { func (e *Engine) triggerAlert(alertID int, title, message string) {
@@ -590,57 +489,11 @@ func (e *Engine) triggerAlert(alertID int, title, message string) {
defer cancel() defer cancel()
if err := provider.Send(ctx, title, message); err != nil { if err := provider.Send(ctx, title, message); err != nil {
e.AddLog(fmt.Sprintf("Alert send failed (%s): %v", cfg.Name, err)) e.AddLog(fmt.Sprintf("Alert send failed (%s): %v", cfg.Name, err))
e.recordAlertResult(alertID, false, err.Error())
} else {
e.recordAlertResult(alertID, true, "")
} }
}() }()
} }
} }
func (e *Engine) recordAlertResult(alertID int, ok bool, errMsg string) {
e.alertHealthMu.Lock()
defer e.alertHealthMu.Unlock()
h := e.alertHealth[alertID]
h.LastSendAt = time.Now()
h.LastSendOK = ok
h.SendCount++
if ok {
h.LastError = ""
} else {
h.LastError = errMsg
h.FailCount++
}
e.alertHealth[alertID] = h
}
func (e *Engine) GetAlertHealth(alertID int) AlertHealth {
e.alertHealthMu.RLock()
defer e.alertHealthMu.RUnlock()
return e.alertHealth[alertID]
}
func (e *Engine) TestAlert(alertID int) error {
cfg, err := e.db.GetAlert(alertID)
if err != nil {
return fmt.Errorf("failed to load alert: %w", err)
}
provider := alert.GetProvider(cfg)
if provider == nil {
return fmt.Errorf("no provider for type %q", cfg.Type)
}
ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
defer cancel()
err = provider.Send(ctx, "🧪 Test Alert", fmt.Sprintf("Test notification from uptop for channel '%s'.", cfg.Name))
if err != nil {
e.recordAlertResult(alertID, false, err.Error())
return err
}
e.recordAlertResult(alertID, true, "")
e.AddLog(fmt.Sprintf("Test alert sent to '%s'", cfg.Name))
return nil
}
func (e *Engine) isInMaintenance(monitorID int) bool { func (e *Engine) isInMaintenance(monitorID int) bool {
inMaint, err := e.db.IsMonitorInMaintenance(monitorID) inMaint, err := e.db.IsMonitorInMaintenance(monitorID)
if err != nil { if err != nil {
@@ -701,17 +554,16 @@ func (e *Engine) SetAggStrategy(strategy AggregationStrategy) {
e.aggStrategy = strategy e.aggStrategy = strategy
} }
func (e *Engine) IngestProbeResult(nodeID string, siteID int, latencyNs int64, isUp bool, errorReason string) { func (e *Engine) IngestProbeResult(nodeID string, siteID int, latencyNs int64, isUp bool) {
e.probeResultsMu.Lock() e.probeResultsMu.Lock()
if e.probeResults[siteID] == nil { if e.probeResults[siteID] == nil {
e.probeResults[siteID] = make(map[string]NodeResult) e.probeResults[siteID] = make(map[string]NodeResult)
} }
e.probeResults[siteID][nodeID] = NodeResult{ e.probeResults[siteID][nodeID] = NodeResult{
NodeID: nodeID, NodeID: nodeID,
IsUp: isUp, IsUp: isUp,
LatencyNs: latencyNs, LatencyNs: latencyNs,
CheckedAt: time.Now(), CheckedAt: time.Now(),
ErrorReason: errorReason,
} }
results := make([]NodeResult, 0, len(e.probeResults[siteID])) results := make([]NodeResult, 0, len(e.probeResults[siteID]))
for _, r := range e.probeResults[siteID] { for _, r := range e.probeResults[siteID] {
@@ -736,7 +588,7 @@ func (e *Engine) IngestProbeResult(nodeID string, siteID int, latencyNs int64, i
updatedSite := site updatedSite := site
updatedSite.Latency = time.Duration(avgLatency) updatedSite.Latency = time.Duration(avgLatency)
updatedSite.LastCheck = time.Now() updatedSite.LastCheck = time.Now()
e.handleStatusChange(updatedSite, rawStatus, 0, time.Duration(avgLatency), errorReason) e.handleStatusChange(updatedSite, rawStatus, 0, time.Duration(avgLatency))
} }
func (e *Engine) GetProbeResults(siteID int) map[string]NodeResult { func (e *Engine) GetProbeResults(siteID int) map[string]NodeResult {
@@ -749,11 +601,3 @@ func (e *Engine) GetProbeResults(siteID int) map[string]NodeResult {
} }
return cp return cp
} }
func (e *Engine) GetStateChanges(siteID int, limit int) []models.StateChange {
changes, err := e.db.GetStateChanges(siteID, limit)
if err != nil {
return nil
}
return changes
}
+27 -48
View File
@@ -2,11 +2,10 @@ package monitor
import ( import (
"fmt" "fmt"
"gitea.lerkolabs.com/lerko/uptop/internal/models"
"sync" "sync"
"testing" "testing"
"time" "time"
"gitea.lerkolabs.com/lerko/uptop/internal/models"
) )
// --- Mock Store --- // --- Mock Store ---
@@ -69,14 +68,12 @@ func (m *mockStore) GetActiveMaintenanceWindows() ([]models.MaintenanceWindow, e
func (m *mockStore) GetAllMaintenanceWindows(int) ([]models.MaintenanceWindow, error) { func (m *mockStore) GetAllMaintenanceWindows(int) ([]models.MaintenanceWindow, error) {
return nil, nil return nil, nil
} }
func (m *mockStore) AddMaintenanceWindow(models.MaintenanceWindow) error { return nil } func (m *mockStore) AddMaintenanceWindow(models.MaintenanceWindow) error { return nil }
func (m *mockStore) EndMaintenanceWindow(int) error { return nil } func (m *mockStore) EndMaintenanceWindow(int) error { return nil }
func (m *mockStore) DeleteMaintenanceWindow(int) error { return nil } func (m *mockStore) DeleteMaintenanceWindow(int) error { return nil }
func (m *mockStore) GetPreference(string) (string, error) { return "", nil } func (m *mockStore) GetPreference(string) (string, error) { return "", nil }
func (m *mockStore) SetPreference(string, string) error { return nil } func (m *mockStore) SetPreference(string, string) error { return nil }
func (m *mockStore) SaveStateChange(int, string, string, string) error { return nil } func (m *mockStore) Close() error { return nil }
func (m *mockStore) GetStateChanges(int, int) ([]models.StateChange, error) { return nil, nil }
func (m *mockStore) Close() error { return nil }
func (m *mockStore) GetAllAlerts() ([]models.AlertConfig, error) { func (m *mockStore) GetAllAlerts() ([]models.AlertConfig, error) {
m.mu.Lock() m.mu.Lock()
@@ -177,7 +174,7 @@ func TestHandleStatusChange_PendingToUp(t *testing.T) {
site := models.Site{ID: 1, Name: "test", Status: "PENDING", MaxRetries: 3, AlertID: 1} site := models.Site{ID: 1, Name: "test", Status: "PENDING", MaxRetries: 3, AlertID: 1}
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "UP", 200, 10*time.Millisecond, "") e.handleStatusChange(site, "UP", 200, 10*time.Millisecond)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if s.Status != "UP" { if s.Status != "UP" {
@@ -198,7 +195,7 @@ func TestHandleStatusChange_UpIncrementFailure(t *testing.T) {
site := models.Site{ID: 1, Name: "test", Status: "UP", MaxRetries: 3, FailureCount: 0} site := models.Site{ID: 1, Name: "test", Status: "UP", MaxRetries: 3, FailureCount: 0}
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "DOWN", 500, 0, "test error") e.handleStatusChange(site, "DOWN", 500, 0)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if s.Status != "UP" { if s.Status != "UP" {
@@ -216,7 +213,7 @@ func TestHandleStatusChange_UpToDown_ExceedsRetries(t *testing.T) {
site := models.Site{ID: 1, Name: "test", Status: "UP", MaxRetries: 2, FailureCount: 2, AlertID: 1} site := models.Site{ID: 1, Name: "test", Status: "UP", MaxRetries: 2, FailureCount: 2, AlertID: 1}
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "DOWN", 500, 0, "test error") e.handleStatusChange(site, "DOWN", 500, 0)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if s.Status != "DOWN" { if s.Status != "DOWN" {
@@ -239,7 +236,7 @@ func TestHandleStatusChange_UpToDown_ZeroRetries(t *testing.T) {
site := models.Site{ID: 1, Name: "test", Status: "UP", MaxRetries: 0, FailureCount: 0, AlertID: 1} site := models.Site{ID: 1, Name: "test", Status: "UP", MaxRetries: 0, FailureCount: 0, AlertID: 1}
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "DOWN", 0, 0, "test error") e.handleStatusChange(site, "DOWN", 0, 0)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if s.Status != "DOWN" { if s.Status != "DOWN" {
@@ -258,7 +255,7 @@ func TestHandleStatusChange_DownToUp_Recovery(t *testing.T) {
site := models.Site{ID: 1, Name: "test", Status: "DOWN", FailureCount: 4, AlertID: 1} site := models.Site{ID: 1, Name: "test", Status: "DOWN", FailureCount: 4, AlertID: 1}
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "UP", 200, 5*time.Millisecond, "") e.handleStatusChange(site, "UP", 200, 5*time.Millisecond)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if s.Status != "UP" { if s.Status != "UP" {
@@ -279,7 +276,7 @@ func TestHandleStatusChange_DownStaysDown(t *testing.T) {
site := models.Site{ID: 1, Name: "test", Status: "DOWN", MaxRetries: 2, FailureCount: 3} site := models.Site{ID: 1, Name: "test", Status: "DOWN", MaxRetries: 2, FailureCount: 3}
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "DOWN", 0, 0, "test error") e.handleStatusChange(site, "DOWN", 0, 0)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if s.Status != "DOWN" { if s.Status != "DOWN" {
@@ -298,7 +295,7 @@ func TestHandleStatusChange_SSLExpired(t *testing.T) {
site := models.Site{ID: 1, Name: "test", Status: "UP", MaxRetries: 0, AlertID: 1} site := models.Site{ID: 1, Name: "test", Status: "UP", MaxRetries: 0, AlertID: 1}
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "SSL EXP", 0, 0, "SSL certificate expired") e.handleStatusChange(site, "SSL EXP", 0, 0)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if s.Status != "SSL EXP" { if s.Status != "SSL EXP" {
@@ -318,7 +315,7 @@ func TestHandleStatusChange_AlertSuppressedMaintenance(t *testing.T) {
site := models.Site{ID: 1, Name: "test", Status: "UP", MaxRetries: 0, AlertID: 1} site := models.Site{ID: 1, Name: "test", Status: "UP", MaxRetries: 0, AlertID: 1}
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "DOWN", 0, 0, "test error") e.handleStatusChange(site, "DOWN", 0, 0)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if s.Status != "DOWN" { if s.Status != "DOWN" {
@@ -349,7 +346,7 @@ func TestHandleStatusChange_RecoverySuppressedMaintenance(t *testing.T) {
site := models.Site{ID: 1, Name: "test", Status: "DOWN", AlertID: 1} site := models.Site{ID: 1, Name: "test", Status: "DOWN", AlertID: 1}
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "UP", 200, 0, "") e.handleStatusChange(site, "UP", 200, 0)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if s.Status != "UP" { if s.Status != "UP" {
@@ -373,7 +370,7 @@ func TestHandleStatusChange_SSLWarning(t *testing.T) {
} }
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "UP", 200, 0, "") e.handleStatusChange(site, "UP", 200, 0)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if !s.SentSSLWarning { if !s.SentSSLWarning {
@@ -396,7 +393,7 @@ func TestHandleStatusChange_SSLWarningNotRepeated(t *testing.T) {
} }
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "UP", 200, 0, "") e.handleStatusChange(site, "UP", 200, 0)
waitAsync() waitAsync()
if len(ms.getAlertCallsSnapshot()) != 0 { if len(ms.getAlertCallsSnapshot()) != 0 {
@@ -415,7 +412,7 @@ func TestHandleStatusChange_SSLWarningReset(t *testing.T) {
} }
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "UP", 200, 0, "") e.handleStatusChange(site, "UP", 200, 0)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if s.SentSSLWarning { if s.SentSSLWarning {
@@ -436,7 +433,7 @@ func TestHandleStatusChange_SSLWarningSuppressedMaint(t *testing.T) {
} }
injectSite(e, site) injectSite(e, site)
e.handleStatusChange(site, "UP", 200, 0, "") e.handleStatusChange(site, "UP", 200, 0)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if !s.SentSSLWarning { if !s.SentSSLWarning {
@@ -455,7 +452,7 @@ func TestHandleStatusChange_InactiveEngine(t *testing.T) {
injectSite(e, site) injectSite(e, site)
e.SetActive(false) e.SetActive(false)
e.handleStatusChange(site, "DOWN", 0, 0, "test error") e.handleStatusChange(site, "DOWN", 0, 0)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if s.Status != "UP" { if s.Status != "UP" {
@@ -537,7 +534,7 @@ func TestCheckPush_DeadlineMissed(t *testing.T) {
site := models.Site{ site := models.Site{
ID: 1, Name: "push", Type: "push", Status: "UP", ID: 1, Name: "push", Type: "push", Status: "UP",
Interval: 10, MaxRetries: 0, Interval: 10, MaxRetries: 0,
LastCheck: time.Now().Add(-120 * time.Second), LastCheck: time.Now().Add(-20 * time.Second),
} }
injectSite(e, site) injectSite(e, site)
@@ -549,24 +546,6 @@ func TestCheckPush_DeadlineMissed(t *testing.T) {
} }
} }
func TestCheckPush_OverdueBecomesLate(t *testing.T) {
ms := newMockStore()
e := newTestEngine(ms)
site := models.Site{
ID: 1, Name: "push", Type: "push", Status: "UP",
Interval: 300,
LastCheck: time.Now().Add(-310 * time.Second),
}
injectSite(e, site)
e.checkPush(site)
s, _ := getSite(e, 1)
if s.Status != "LATE" {
t.Errorf("expected LATE when overdue but within grace, got %s", s.Status)
}
}
func TestCheckPush_WithinDeadline(t *testing.T) { func TestCheckPush_WithinDeadline(t *testing.T) {
ms := newMockStore() ms := newMockStore()
e := newTestEngine(ms) e := newTestEngine(ms)
@@ -584,20 +563,20 @@ func TestCheckPush_WithinDeadline(t *testing.T) {
} }
} }
func TestCheckPush_PendingStaysPending(t *testing.T) { func TestCheckPush_PendingToUp(t *testing.T) {
ms := newMockStore() ms := newMockStore()
e := newTestEngine(ms) e := newTestEngine(ms)
site := models.Site{ site := models.Site{
ID: 1, Name: "push", Type: "push", Status: "PENDING", ID: 1, Name: "push", Type: "push", Status: "PENDING",
Interval: 60, Interval: 60, LastCheck: time.Now(),
} }
injectSite(e, site) injectSite(e, site)
e.checkPush(site) e.checkPush(site)
s, _ := getSite(e, 1) s, _ := getSite(e, 1)
if s.Status != "PENDING" { if s.Status != "UP" {
t.Errorf("expected PENDING to stay until first heartbeat, got %s", s.Status) t.Errorf("expected UP, got %s", s.Status)
} }
} }
@@ -1012,7 +991,7 @@ func TestConcurrent_HandleStatusChangeAndGetState(t *testing.T) {
wg.Add(2) wg.Add(2)
go func() { go func() {
defer wg.Done() defer wg.Done()
e.handleStatusChange(site, "DOWN", 500, 0, "test error") e.handleStatusChange(site, "DOWN", 500, 0)
}() }()
go func() { go func() {
defer wg.Done() defer wg.Done()
+4 -6
View File
@@ -67,7 +67,6 @@ var statusTpl = template.Must(template.New("status").Parse(`
.UP { background: #9ece6a; color: #1a1b26; } .UP { background: #9ece6a; color: #1a1b26; }
.DOWN { background: #f7768e; color: #1a1b26; } .DOWN { background: #f7768e; color: #1a1b26; }
.PENDING { background: #e0af68; color: #1a1b26; } .PENDING { background: #e0af68; color: #1a1b26; }
.LATE { background: #e0af68; color: #1a1b26; }
.SSL-EXP { background: #e0af68; color: #1a1b26; } .SSL-EXP { background: #e0af68; color: #1a1b26; }
.PAUSED { background: #565f89; color: #c0caf5; } .PAUSED { background: #565f89; color: #c0caf5; }
.MAINT { background: #bb9af7; color: #1a1b26; } .MAINT { background: #bb9af7; color: #1a1b26; }
@@ -404,10 +403,9 @@ func Start(cfg ServerConfig, s store.Store, eng *monitor.Engine) *http.Server {
var req struct { var req struct {
NodeID string `json:"node_id"` NodeID string `json:"node_id"`
Results []struct { Results []struct {
SiteID int `json:"site_id"` SiteID int `json:"site_id"`
LatencyNs int64 `json:"latency_ns"` LatencyNs int64 `json:"latency_ns"`
IsUp bool `json:"is_up"` IsUp bool `json:"is_up"`
ErrorReason string `json:"error_reason"`
} `json:"results"` } `json:"results"`
} }
if err := json.NewDecoder(r.Body).Decode(&req); err != nil { if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
@@ -422,7 +420,7 @@ func Start(cfg ServerConfig, s store.Store, eng *monitor.Engine) *http.Server {
if err := s.SaveCheckFromNode(result.SiteID, req.NodeID, result.LatencyNs, result.IsUp); err != nil { if err := s.SaveCheckFromNode(result.SiteID, req.NodeID, result.LatencyNs, result.IsUp); err != nil {
log.Printf("Failed to save probe result: %v", err) log.Printf("Failed to save probe result: %v", err)
} }
eng.IngestProbeResult(req.NodeID, result.SiteID, result.LatencyNs, result.IsUp, result.ErrorReason) eng.IngestProbeResult(req.NodeID, result.SiteID, result.LatencyNs, result.IsUp)
} }
if err := s.UpdateNodeLastSeen(req.NodeID); err != nil { if err := s.UpdateNodeLastSeen(req.NodeID); err != nil {
log.Printf("Failed to update node last seen: %v", err) log.Printf("Failed to update node last seen: %v", err)
+9 -12
View File
@@ -4,14 +4,13 @@ import (
"bytes" "bytes"
"encoding/json" "encoding/json"
"fmt" "fmt"
"gitea.lerkolabs.com/lerko/uptop/internal/models"
"gitea.lerkolabs.com/lerko/uptop/internal/monitor"
"net" "net"
"net/http" "net/http"
"sync" "sync"
"testing" "testing"
"time" "time"
"gitea.lerkolabs.com/lerko/uptop/internal/models"
"gitea.lerkolabs.com/lerko/uptop/internal/monitor"
) )
// --- Mock Store --- // --- Mock Store ---
@@ -70,15 +69,13 @@ func (m *mockStore) LoadLogs(int) ([]string, error) { return nil, nil
func (m *mockStore) GetAllMaintenanceWindows(int) ([]models.MaintenanceWindow, error) { func (m *mockStore) GetAllMaintenanceWindows(int) ([]models.MaintenanceWindow, error) {
return nil, nil return nil, nil
} }
func (m *mockStore) AddMaintenanceWindow(models.MaintenanceWindow) error { return nil } func (m *mockStore) AddMaintenanceWindow(models.MaintenanceWindow) error { return nil }
func (m *mockStore) EndMaintenanceWindow(int) error { return nil } func (m *mockStore) EndMaintenanceWindow(int) error { return nil }
func (m *mockStore) DeleteMaintenanceWindow(int) error { return nil } func (m *mockStore) DeleteMaintenanceWindow(int) error { return nil }
func (m *mockStore) IsMonitorInMaintenance(int) (bool, error) { return false, nil } func (m *mockStore) IsMonitorInMaintenance(int) (bool, error) { return false, nil }
func (m *mockStore) GetPreference(string) (string, error) { return "", nil } func (m *mockStore) GetPreference(string) (string, error) { return "", nil }
func (m *mockStore) SetPreference(string, string) error { return nil } func (m *mockStore) SetPreference(string, string) error { return nil }
func (m *mockStore) SaveStateChange(int, string, string, string) error { return nil } func (m *mockStore) Close() error { return nil }
func (m *mockStore) GetStateChanges(int, int) ([]models.StateChange, error) { return nil, nil }
func (m *mockStore) Close() error { return nil }
func (m *mockStore) ExportData() (models.Backup, error) { func (m *mockStore) ExportData() (models.Backup, error) {
return models.Backup{ return models.Backup{
-9
View File
@@ -72,15 +72,6 @@ func (d *PostgresDialect) CreateTablesSQL() []string {
key TEXT PRIMARY KEY, key TEXT PRIMARY KEY,
value TEXT NOT NULL value TEXT NOT NULL
)`, )`,
`CREATE TABLE IF NOT EXISTS state_changes (
id SERIAL PRIMARY KEY,
site_id INTEGER NOT NULL,
from_status TEXT NOT NULL,
to_status TEXT NOT NULL,
error_reason TEXT DEFAULT '',
changed_at TIMESTAMP DEFAULT NOW()
)`,
`CREATE INDEX IF NOT EXISTS idx_state_changes_site ON state_changes(site_id, changed_at DESC)`,
} }
} }
-9
View File
@@ -79,15 +79,6 @@ func (d *SQLiteDialect) CreateTablesSQL() []string {
key TEXT PRIMARY KEY, key TEXT PRIMARY KEY,
value TEXT NOT NULL value TEXT NOT NULL
)`, )`,
`CREATE TABLE IF NOT EXISTS state_changes (
id INTEGER PRIMARY KEY AUTOINCREMENT,
site_id INTEGER NOT NULL,
from_status TEXT NOT NULL,
to_status TEXT NOT NULL,
error_reason TEXT DEFAULT '',
changed_at DATETIME DEFAULT CURRENT_TIMESTAMP
)`,
`CREATE INDEX IF NOT EXISTS idx_state_changes_site ON state_changes(site_id, changed_at DESC)`,
} }
} }
-23
View File
@@ -347,29 +347,6 @@ func (s *SQLStore) DeleteUser(id int) error {
return err return err
} }
func (s *SQLStore) SaveStateChange(siteID int, fromStatus, toStatus, errorReason string) error {
_, err := s.db.Exec(s.q("INSERT INTO state_changes (site_id, from_status, to_status, error_reason) VALUES (?, ?, ?, ?)"),
siteID, fromStatus, toStatus, errorReason)
return err
}
func (s *SQLStore) GetStateChanges(siteID int, limit int) ([]models.StateChange, error) {
rows, err := s.db.Query(s.q("SELECT id, site_id, from_status, to_status, error_reason, changed_at FROM state_changes WHERE site_id = ? ORDER BY changed_at DESC LIMIT ?"), siteID, limit)
if err != nil {
return nil, err
}
defer rows.Close()
var changes []models.StateChange
for rows.Next() {
var sc models.StateChange
if err := rows.Scan(&sc.ID, &sc.SiteID, &sc.FromStatus, &sc.ToStatus, &sc.ErrorReason, &sc.ChangedAt); err != nil {
return changes, err
}
changes = append(changes, sc)
}
return changes, rows.Err()
}
func (s *SQLStore) SaveCheck(siteID int, latencyNs int64, isUp bool) error { func (s *SQLStore) SaveCheck(siteID int, latencyNs int64, isUp bool) error {
return s.SaveCheckFromNode(siteID, "", latencyNs, isUp) return s.SaveCheckFromNode(siteID, "", latencyNs, isUp)
} }
-4
View File
@@ -38,10 +38,6 @@ type Store interface {
SaveCheckFromNode(siteID int, nodeID string, latencyNs int64, isUp bool) error SaveCheckFromNode(siteID int, nodeID string, latencyNs int64, isUp bool) error
LoadAllHistory(limit int) (map[int][]models.CheckRecord, error) LoadAllHistory(limit int) (map[int][]models.CheckRecord, error)
// State Changes
SaveStateChange(siteID int, fromStatus, toStatus, errorReason string) error
GetStateChanges(siteID int, limit int) ([]models.StateChange, error)
// Nodes // Nodes
RegisterNode(node models.ProbeNode) error RegisterNode(node models.ProbeNode) error
GetNode(id string) (models.ProbeNode, error) GetNode(id string) (models.ProbeNode, error)
+5 -96
View File
@@ -2,10 +2,7 @@ package tui
import ( import (
"fmt" "fmt"
"strings"
"time"
"gitea.lerkolabs.com/lerko/uptop/internal/monitor"
tea "github.com/charmbracelet/bubbletea" tea "github.com/charmbracelet/bubbletea"
"github.com/charmbracelet/huh" "github.com/charmbracelet/huh"
"github.com/charmbracelet/lipgloss" "github.com/charmbracelet/lipgloss"
@@ -116,122 +113,34 @@ func fmtAlertConfig(alert struct {
} }
} }
func fmtAlertHealth(h monitor.AlertHealth) string {
if h.LastSendAt.IsZero() {
return subtleStyle.Render("●")
}
if h.LastSendOK {
return specialStyle.Render("●")
}
return dangerStyle.Render("●")
}
func fmtAlertLastSent(h monitor.AlertHealth) string {
if h.LastSendAt.IsZero() {
return subtleStyle.Render("never")
}
d := time.Since(h.LastSendAt)
if d < time.Minute {
return fmt.Sprintf("%ds ago", int(d.Seconds()))
}
if d < time.Hour {
return fmt.Sprintf("%dm ago", int(d.Minutes()))
}
if d < 24*time.Hour {
return fmt.Sprintf("%dh ago", int(d.Hours()))
}
return fmt.Sprintf("%dd ago", int(d.Hours())/24)
}
func (m Model) viewAlertsTab() string { func (m Model) viewAlertsTab() string {
if len(m.alerts) == 0 { if len(m.alerts) == 0 {
return "\n No alert channels configured. Press [n] to add one." return "\n No alert channels configured. Press [n] to add one."
} }
var headers []string
var widths []int
if m.isWide() {
headers = []string{"#", "", "NAME", "TYPE", "CONFIG", "LAST SENT"}
widths = []int{4, 3, 18, 12, 40, 12}
} else {
headers = []string{"#", "", "NAME", "TYPE", "CONFIG", "SENT"}
widths = []int{4, 3, 14, 10, 24, 8}
}
nameW := widths[2]
cfgW := widths[4]
return m.renderTable( return m.renderTable(
headers, []string{"#", "NAME", "TYPE", "CONFIG"},
len(m.alerts), len(m.alerts),
func(start, end int) [][]string { func(start, end int) [][]string {
var rows [][]string var rows [][]string
for i := start; i < end; i++ { for i := start; i < end; i++ {
a := m.alerts[i] a := m.alerts[i]
h := m.engine.GetAlertHealth(a.ID)
rows = append(rows, []string{ rows = append(rows, []string{
fmt.Sprintf("%d", i+1), fmt.Sprintf("%d", i+1),
fmtAlertHealth(h), m.zones.Mark(fmt.Sprintf("alert-%d", i), limitStr(a.Name, 15)),
m.zones.Mark(fmt.Sprintf("alert-%d", i), limitStr(a.Name, nameW-2)),
fmtAlertType(a.Type), fmtAlertType(a.Type),
limitStr(fmtAlertConfig(struct { fmtAlertConfig(struct {
Type string Type string
Settings map[string]string Settings map[string]string
}{a.Type, a.Settings}), cfgW-2), }{a.Type, a.Settings}),
fmtAlertLastSent(h),
}) })
} }
return rows return rows
}, },
widths, nil, nil, nil,
) )
} }
func (m Model) viewAlertDetailPanel() string {
if m.cursor >= len(m.alerts) {
return ""
}
a := m.alerts[m.cursor]
h := m.engine.GetAlertHealth(a.ID)
var b strings.Builder
b.WriteString(subtleStyle.Render(" Alerts > ") + titleStyle.Render(a.Name) + "\n\n")
row := func(label, value string) {
fmt.Fprintf(&b, " %-16s %s\n", subtleStyle.Render(label), value)
}
row("Type", fmtAlertType(a.Type))
if h.LastSendAt.IsZero() {
row("Health", subtleStyle.Render("never sent"))
} else if h.LastSendOK {
row("Health", specialStyle.Render("OK"))
} else {
row("Health", dangerStyle.Render("FAILED"))
}
if !h.LastSendAt.IsZero() {
row("Last Sent", h.LastSendAt.Format("2006-01-02 15:04:05")+" ("+fmtAlertLastSent(h)+")")
}
if h.SendCount > 0 {
row("Sends", fmt.Sprintf("%d sent, %d failed", h.SendCount, h.FailCount))
}
if h.LastError != "" {
row("Last Error", dangerStyle.Render(limitStr(h.LastError, 60)))
}
b.WriteString("\n" + subtleStyle.Render(" CONFIGURATION") + "\n")
for k, v := range a.Settings {
row(k, v)
}
b.WriteString("\n\n")
b.WriteString(subtleStyle.Render(" [i/Esc] Back [e] Edit [t] Test [q] Quit"))
return lipgloss.NewStyle().Padding(1, 2).Render(b.String())
}
func (m *Model) initAlertHuhForm() tea.Cmd { func (m *Model) initAlertHuhForm() tea.Cmd {
m.alertFormData = &alertFormData{ m.alertFormData = &alertFormData{
AlertType: "discord", AlertType: "discord",
+21 -89
View File
@@ -5,83 +5,27 @@ import (
"strings" "strings"
) )
type logSeverity int func colorizeLog(line string) string {
const (
severityInfo logSeverity = iota
severityWarn
severityDown
severityUp
severitySystem
)
func classifyLog(line string) logSeverity {
lower := strings.ToLower(line) lower := strings.ToLower(line)
switch { switch {
case strings.Contains(lower, "confirmed down"), case strings.Contains(lower, "confirmed down"),
strings.Contains(lower, "is down"), strings.Contains(lower, "is down"),
strings.Contains(lower, "missed heartbeat"), strings.Contains(lower, "missed heartbeat"),
strings.Contains(lower, "alert send failed"): strings.Contains(lower, "failed check"),
return severityDown strings.Contains(lower, "ssl warning"):
return dangerStyle.Render(line)
case strings.Contains(lower, "recovered"), case strings.Contains(lower, "recovered"),
strings.Contains(lower, "is up"), strings.Contains(lower, "is up"),
strings.Contains(lower, "recovery"), strings.Contains(lower, "recovery"):
strings.Contains(lower, "first heartbeat"): return specialStyle.Render(line)
return severityUp
case strings.Contains(lower, "failed check"),
strings.Contains(lower, "ssl warning"),
strings.Contains(lower, "overdue"),
strings.Contains(lower, "was late"):
return severityWarn
case strings.Contains(lower, "engine"), case strings.Contains(lower, "engine"),
strings.Contains(lower, "cluster"), strings.Contains(lower, "cluster"):
strings.Contains(lower, "loaded"), return titleStyle.Render(line)
strings.Contains(lower, "paused"),
strings.Contains(lower, "resumed"):
return severitySystem
default: default:
return severityInfo return line
} }
} }
func isImportantLog(sev logSeverity) bool {
return sev == severityDown || sev == severityUp || sev == severitySystem
}
func renderLogTag(sev logSeverity) string {
switch sev {
case severityDown:
return dangerStyle.Render(" DOWN ")
case severityUp:
return specialStyle.Render(" UP ")
case severityWarn:
return warnStyle.Render(" WARN ")
case severitySystem:
return titleStyle.Render(" SYS ")
default:
return subtleStyle.Render(" info ")
}
}
func renderLogLine(line string) string {
sev := classifyLog(line)
tag := renderLogTag(sev)
ts := ""
msg := line
if len(line) > 10 && line[0] == '[' {
if idx := strings.Index(line, "]"); idx > 0 && idx < 12 {
ts = subtleStyle.Render(line[1:idx])
msg = strings.TrimSpace(line[idx+1:])
}
}
if ts != "" {
return fmt.Sprintf(" %s %s %s", ts, tag, msg)
}
return fmt.Sprintf(" %s %s", tag, msg)
}
func (m Model) viewLogsTab() string { func (m Model) viewLogsTab() string {
content := m.logViewport.View() content := m.logViewport.View()
if strings.TrimSpace(content) == "" || content == "Waiting for logs..." { if strings.TrimSpace(content) == "" || content == "Waiting for logs..." {
@@ -89,34 +33,22 @@ func (m Model) viewLogsTab() string {
} }
lines := strings.Split(content, "\n") lines := strings.Split(content, "\n")
var rendered []string var colored []string
total := 0
shown := 0
for _, line := range lines { for _, line := range lines {
if strings.TrimSpace(line) == "" { if line == "" {
colored = append(colored, line)
continue continue
} }
total++ colored = append(colored, colorizeLog(line))
sev := classifyLog(line) }
if m.logFilterImportant && !isImportantLog(sev) {
continue count := 0
for _, l := range lines {
if strings.TrimSpace(l) != "" {
count++
} }
shown++
rendered = append(rendered, renderLogLine(line))
} }
filterLabel := "All" header := subtleStyle.Render(fmt.Sprintf(" %d entries [↑/↓] Scroll [PgUp/PgDn] Page", count))
if m.logFilterImportant { return "\n" + header + "\n\n" + strings.Join(colored, "\n")
filterLabel = "Important"
}
header := subtleStyle.Render(fmt.Sprintf(
" %d entries [↑/↓] Scroll [PgUp/PgDn] Page [f] Filter: %s", shown, filterLabel))
if m.logFilterImportant && shown < total {
header += subtleStyle.Render(fmt.Sprintf(" (%d hidden)", total-shown))
}
return "\n" + header + "\n\n" + strings.Join(rendered, "\n")
} }
+11 -28
View File
@@ -2,11 +2,10 @@ package tui
import ( import (
"fmt" "fmt"
"gitea.lerkolabs.com/lerko/uptop/internal/models"
"strconv" "strconv"
"time" "time"
"gitea.lerkolabs.com/lerko/uptop/internal/models"
tea "github.com/charmbracelet/bubbletea" tea "github.com/charmbracelet/bubbletea"
"github.com/charmbracelet/huh" "github.com/charmbracelet/huh"
"github.com/charmbracelet/lipgloss" "github.com/charmbracelet/lipgloss"
@@ -41,19 +40,19 @@ func fmtMaintType(t string) string {
return maintStyle.Render("maintenance") return maintStyle.Render("maintenance")
} }
func fmtMaintMonitorW(monitorID int, sites []models.Site, maxW int) string { func fmtMaintMonitor(monitorID int, sites []models.Site) string {
if monitorID == 0 { if monitorID == 0 {
return "All" return "All"
} }
for _, s := range sites { for _, s := range sites {
if s.ID == monitorID { if s.ID == monitorID {
return limitStr(s.Name, maxW) return limitStr(s.Name, 18)
} }
} }
return fmt.Sprintf("#%d", monitorID) return fmt.Sprintf("#%d", monitorID)
} }
func fmtMaintTime(t time.Time, colW int) string { func fmtMaintTime(t time.Time) string {
if t.IsZero() { if t.IsZero() {
return subtleStyle.Render("—") return subtleStyle.Render("—")
} }
@@ -61,10 +60,7 @@ func fmtMaintTime(t time.Time, colW int) string {
if t.Year() == now.Year() && t.YearDay() == now.YearDay() { if t.Year() == now.Year() && t.YearDay() == now.YearDay() {
return t.Format("15:04") return t.Format("15:04")
} }
if colW >= 14 { return t.Format("15:04 Jan 02")
return t.Format("15:04 Jan 02")
}
return t.Format("Jan 02")
} }
func (m Model) isMonitorInMaintenance(monitorID int) bool { func (m Model) isMonitorInMaintenance(monitorID int) bool {
@@ -96,21 +92,8 @@ func (m Model) viewMaintTab() string {
return "\n No maintenance windows or incidents. Press [n] to create one." return "\n No maintenance windows or incidents. Press [n] to create one."
} }
var headers []string
var widths []int
if m.isWide() {
headers = []string{"#", "TITLE", "TYPE", "MONITORS", "STATUS", "STARTED", "ENDS"}
widths = []int{4, 24, 14, 22, 12, 16, 16}
} else {
headers = []string{"#", "TITLE", "TYPE", "MON", "ST", "START", "ENDS"}
widths = []int{4, 14, 13, 14, 11, 14, 14}
}
titleW := widths[1]
monW := widths[3]
timeW := widths[5]
return m.renderTable( return m.renderTable(
headers, []string{"#", "TITLE", "TYPE", "MONITORS", "STATUS", "STARTED", "ENDS"},
len(m.maintenanceWindows), len(m.maintenanceWindows),
func(start, end int) [][]string { func(start, end int) [][]string {
var rows [][]string var rows [][]string
@@ -119,17 +102,17 @@ func (m Model) viewMaintTab() string {
mw := m.maintenanceWindows[i] mw := m.maintenanceWindows[i]
rows = append(rows, []string{ rows = append(rows, []string{
strconv.Itoa(i + 1), strconv.Itoa(i + 1),
m.zones.Mark(fmt.Sprintf("maint-%d", i), limitStr(mw.Title, titleW-2)), m.zones.Mark(fmt.Sprintf("maint-%d", i), limitStr(mw.Title, 24)),
fmtMaintType(mw.Type), fmtMaintType(mw.Type),
fmtMaintMonitorW(mw.MonitorID, allSites, monW-2), fmtMaintMonitor(mw.MonitorID, allSites),
fmtMaintStatus(mw), fmtMaintStatus(mw),
fmtMaintTime(mw.StartTime, timeW), fmtMaintTime(mw.StartTime),
fmtMaintTime(mw.EndTime, timeW), fmtMaintTime(mw.EndTime),
}) })
} }
return rows return rows
}, },
widths, []int{6, 0, 14, 20, 12, 16, 16},
nil, nil,
) )
} }
+4 -13
View File
@@ -10,25 +10,16 @@ func (m Model) viewNodesTab() string {
return "\n No probe nodes connected." return "\n No probe nodes connected."
} }
var headers []string colWidths := []int{0, 12, 20, 10, 8}
var widths []int
if m.isWide() {
headers = []string{"NAME", "REGION", "LAST SEEN", "VERSION", "STATUS"}
widths = []int{24, 14, 16, 12, 10}
} else {
headers = []string{"NAME", "REGION", "SEEN", "VER", "STATUS"}
widths = []int{16, 10, 10, 8, 8}
}
nameW := widths[0]
return m.renderTable( return m.renderTable(
headers, []string{"NAME", "REGION", "LAST SEEN", "VERSION", "STATUS"},
len(m.nodes), len(m.nodes),
func(start, end int) [][]string { func(start, end int) [][]string {
var rows [][]string var rows [][]string
for i := start; i < end; i++ { for i := start; i < end; i++ {
node := m.nodes[i] node := m.nodes[i]
name := limitStr(node.Name, nameW-2) name := limitStr(node.Name, 20)
if name == "" { if name == "" {
name = node.ID name = node.ID
} }
@@ -46,7 +37,7 @@ func (m Model) viewNodesTab() string {
} }
return rows return rows
}, },
widths, colWidths,
nil, nil,
) )
} }
+43 -194
View File
@@ -60,18 +60,14 @@ type siteFormData struct {
Regions string Regions string
} }
func latencySparkline(latencies []time.Duration, statuses []bool, width int) string { func latencySparkline(latencies []time.Duration, width int) string {
if len(latencies) == 0 { if len(latencies) == 0 {
return subtleStyle.Render(strings.Repeat("·", width)) return subtleStyle.Render(strings.Repeat("·", width))
} }
samples := latencies samples := latencies
sampledStatuses := statuses
if len(samples) > width { if len(samples) > width {
samples = samples[len(samples)-width:] samples = samples[len(samples)-width:]
if len(sampledStatuses) > width {
sampledStatuses = sampledStatuses[len(sampledStatuses)-width:]
}
} }
minL, maxL := samples[0], samples[0] minL, maxL := samples[0], samples[0]
@@ -89,7 +85,7 @@ func latencySparkline(latencies []time.Duration, statuses []bool, width int) str
sb.WriteString(subtleStyle.Render(strings.Repeat("·", remaining))) sb.WriteString(subtleStyle.Render(strings.Repeat("·", remaining)))
} }
spread := maxL - minL spread := maxL - minL
for i, l := range samples { for _, l := range samples {
idx := 0 idx := 0
if spread > 0 { if spread > 0 {
idx = int(float64(l-minL) / float64(spread) * 7) idx = int(float64(l-minL) / float64(spread) * 7)
@@ -98,18 +94,13 @@ func latencySparkline(latencies []time.Duration, statuses []bool, width int) str
} }
} }
ch := string(sparkChars[idx]) ch := string(sparkChars[idx])
isDown := i < len(sampledStatuses) && !sampledStatuses[i] ms := l.Milliseconds()
if isDown { if ms < 200 {
sb.WriteString(dangerStyle.Render(ch)) sb.WriteString(specialStyle.Render(ch))
} else if ms < 500 {
sb.WriteString(warnStyle.Render(ch))
} else { } else {
ms := l.Milliseconds() sb.WriteString(dangerStyle.Render(ch))
if ms < 200 {
sb.WriteString(specialStyle.Render(ch))
} else if ms < 500 {
sb.WriteString(warnStyle.Render(ch))
} else {
sb.WriteString(dangerStyle.Render(ch))
}
} }
} }
return sb.String() return sb.String()
@@ -311,8 +302,6 @@ func fmtStatus(status string, paused bool, inMaint bool) string {
switch status { switch status {
case "DOWN", "SSL EXP": case "DOWN", "SSL EXP":
return dangerStyle.Render(status) return dangerStyle.Render(status)
case "LATE":
return warnStyle.Render(status)
case "PENDING": case "PENDING":
return subtleStyle.Render(status) return subtleStyle.Render(status)
default: default:
@@ -320,94 +309,28 @@ func fmtStatus(status string, paused bool, inMaint bool) string {
} }
} }
func fmtDuration(d time.Duration) string { func (m Model) dynamicWidths() (nameW, sparkW int) {
if d < time.Minute { fixed := 6 + 10 + 10 + 8 + 8 + 7 + 9 // #, TYPE, STATUS, LATENCY, UPTIME, SSL, RETRY
return fmt.Sprintf("%ds", int(d.Seconds())) overhead := 30 // cell padding + borders
} avail := m.termWidth - chromePadH - 2 - fixed - overhead
if d < time.Hour { if avail < 30 {
return fmt.Sprintf("%dm", int(d.Minutes())) avail = 30
}
if d < 24*time.Hour {
h := int(d.Hours())
m := int(d.Minutes()) % 60
if m > 0 {
return fmt.Sprintf("%dh %dm", h, m)
}
return fmt.Sprintf("%dh", h)
}
days := int(d.Hours()) / 24
hours := int(d.Hours()) % 24
if hours > 0 {
return fmt.Sprintf("%dd %dh", days, hours)
}
return fmt.Sprintf("%dd", days)
}
type tableLayout struct {
nameW, sparkW int
headers []string
colWidths []int
}
func (m Model) computeLayout() tableLayout {
wide := m.isWide()
var fixed int
var headers []string
var widths []int
if wide {
// # NAME TYPE STATUS LATENCY UPTIME HISTORY SSL RETRIES
headers = []string{"#", "NAME", "TYPE", "STATUS", "LATENCY", "UPTIME", "HISTORY", "SSL", "RETRIES"}
widths = []int{4, 0, 10, 10, 10, 8, 0, 7, 9}
fixed = 4 + 10 + 10 + 10 + 8 + 7 + 9
} else {
// # NAME TYPE STATUS LAT UP% HISTORY SSL RT
headers = []string{"#", "NAME", "TYPE", "STATUS", "LAT", "UP%", "HISTORY", "SSL", "RT"}
widths = []int{4, 0, 8, 8, 7, 8, 0, 5, 5}
fixed = 4 + 8 + 8 + 7 + 8 + 5 + 5
}
numCols := len(headers)
borderOverhead := 2 + (numCols - 1)
avail := m.termWidth - chromePadH - 2 - borderOverhead - fixed
if avail < 20 {
avail = 20
}
maxName := 0
for _, s := range m.sites {
if n := len([]rune(s.Name)); n > maxName {
maxName = n
}
}
maxName += 4
nameW := avail / 2
if nameW > maxName {
nameW = maxName
} }
nameW = avail / 2
sparkW = avail - nameW - 2 // -2 for spark column padding
if nameW < 13 { if nameW < 13 {
nameW = 13 nameW = 13
} }
if nameW > 40 { if nameW > 40 {
nameW = 40 nameW = 40
} }
sparkW := avail - nameW
if sparkW < 10 { if sparkW < 10 {
sparkW = 10 sparkW = 10
} }
if sparkW > 60 {
widths[1] = nameW sparkW = 60
widths[6] = sparkW
return tableLayout{
nameW: nameW,
sparkW: sparkW,
headers: headers,
colWidths: widths,
} }
return
} }
func (m Model) viewSitesTab() string { func (m Model) viewSitesTab() string {
@@ -425,16 +348,12 @@ func (m Model) viewSitesTab() string {
return "\n" + welcome return "\n" + welcome
} }
layout := m.computeLayout() nameW, sparkWidth := m.dynamicWidths()
nameW := layout.nameW colWidths := []int{6, 0, 10, 10, 8, 8, sparkWidth + 2, 7, 9}
sparkWidth := layout.sparkW - 2
if sparkWidth < 8 {
sparkWidth = 8
}
var groupRows map[int]bool var groupRows map[int]bool
return m.renderTable( return m.renderTable(
layout.headers, []string{"#", "NAME", "TYPE", "STATUS", "LATENCY", "UPTIME", "HISTORY", "SSL", "RETRY"},
len(m.sites), len(m.sites),
func(start, end int) [][]string { func(start, end int) [][]string {
groupRows = make(map[int]bool) groupRows = make(map[int]bool)
@@ -447,7 +366,7 @@ func (m Model) viewSitesTab() string {
icon := typeIcon("group", m.collapsed[site.ID]) icon := typeIcon("group", m.collapsed[site.ID])
rows = append(rows, []string{ rows = append(rows, []string{
strconv.Itoa(i + 1), strconv.Itoa(i + 1),
m.zones.Mark(fmt.Sprintf("site-%d", i), icon+" "+limitStr(site.Name, nameW-4)), m.zones.Mark(fmt.Sprintf("site-%d", i), icon+" "+limitStr(site.Name, nameW-2)),
"group", "group",
fmtStatus(site.Status, site.Paused, m.isMonitorInMaintenance(site.ID)), fmtStatus(site.Status, site.Paused, m.isMonitorInMaintenance(site.ID)),
subtleStyle.Render("—"), subtleStyle.Render("—"),
@@ -465,17 +384,9 @@ func (m Model) viewSitesTab() string {
if i+1 >= len(m.sites) || m.sites[i+1].ParentID != site.ParentID { if i+1 >= len(m.sites) || m.sites[i+1].ParentID != site.ParentID {
prefix = "└" prefix = "└"
} }
name = prefix + " " + limitStr(name, nameW-4) name = prefix + " " + limitStr(name, nameW-2)
} else { } else {
name = limitStr(name, nameW-2) name = limitStr(name, nameW)
}
if (site.Status == "DOWN" || site.Status == "SSL EXP" || site.Status == "LATE") && site.LastError != "" {
nameLen := len([]rune(name))
errSpace := nameW - nameLen - 3
if errSpace > 10 {
name = name + " " + subtleStyle.Render(limitStr(site.LastError, errSpace))
}
} }
hist, _ := m.engine.GetHistory(site.ID) hist, _ := m.engine.GetHistory(site.ID)
@@ -483,7 +394,7 @@ func (m Model) viewSitesTab() string {
if site.Type == "push" { if site.Type == "push" {
spark = heartbeatSparkline(hist.Statuses, sparkWidth) spark = heartbeatSparkline(hist.Statuses, sparkWidth)
} else { } else {
spark = latencySparkline(hist.Latencies, hist.Statuses, sparkWidth) spark = latencySparkline(hist.Latencies, sparkWidth)
} }
rows = append(rows, []string{ rows = append(rows, []string{
@@ -500,7 +411,7 @@ func (m Model) viewSitesTab() string {
} }
return rows return rows
}, },
layout.colWidths, colWidths,
func(row, col int) *lipgloss.Style { func(row, col int) *lipgloss.Style {
if groupRows[row] { if groupRows[row] {
s := siteGroupStyle s := siteGroupStyle
@@ -820,30 +731,7 @@ func (m Model) viewDetailPanel() string {
fmt.Fprintf(&b, " %-16s %s\n", subtleStyle.Render(label), value) fmt.Fprintf(&b, " %-16s %s\n", subtleStyle.Render(label), value)
} }
section := func(label string) {
b.WriteString("\n" + subtleStyle.Render(" "+label) + "\n")
}
row("Status", fmtStatus(site.Status, site.Paused, m.isMonitorInMaintenance(site.ID))) row("Status", fmtStatus(site.Status, site.Paused, m.isMonitorInMaintenance(site.ID)))
if (site.Status == "DOWN" || site.Status == "SSL EXP" || site.Status == "LATE") && site.LastError != "" {
row("Error", dangerStyle.Render(limitStr(site.LastError, 60)))
}
if site.Type == "http" && site.StatusCode > 0 {
row("HTTP Code", strconv.Itoa(site.StatusCode))
}
if !site.StatusChangedAt.IsZero() {
dur := time.Since(site.StatusChangedAt)
row("State Since", site.StatusChangedAt.Format("2006-01-02 15:04:05")+" ("+fmtDuration(dur)+")")
}
if !site.LastSuccessAt.IsZero() {
ago := time.Since(site.LastSuccessAt)
row("Last Success", site.LastSuccessAt.Format("15:04:05")+" ("+fmtDuration(ago)+" ago)")
}
if m.isMonitorInMaintenance(site.ID) { if m.isMonitorInMaintenance(site.ID) {
for _, mw := range m.maintenanceWindows { for _, mw := range m.maintenanceWindows {
if mw.Type == "maintenance" && (mw.MonitorID == 0 || mw.MonitorID == site.ID || mw.MonitorID == site.ParentID) { if mw.Type == "maintenance" && (mw.MonitorID == 0 || mw.MonitorID == site.ID || mw.MonitorID == site.ParentID) {
@@ -852,8 +740,6 @@ func (m Model) viewDetailPanel() string {
} }
} }
} }
section("ENDPOINT")
row("Type", site.Type) row("Type", site.Type)
if site.URL != "" { if site.URL != "" {
row("URL", site.URL) row("URL", site.URL)
@@ -864,45 +750,31 @@ func (m Model) viewDetailPanel() string {
if site.Port > 0 { if site.Port > 0 {
row("Port", strconv.Itoa(site.Port)) row("Port", strconv.Itoa(site.Port))
} }
section("TIMING")
row("Interval", fmt.Sprintf("%ds", site.Interval)) row("Interval", fmt.Sprintf("%ds", site.Interval))
if site.Timeout > 0 { row("Timeout", fmt.Sprintf("%ds", site.Timeout))
row("Timeout", fmt.Sprintf("%ds", site.Timeout))
}
row("Latency", fmtLatency(site.Latency)) row("Latency", fmtLatency(site.Latency))
row("Uptime", fmtUptime(hist.Statuses)) row("Uptime", fmtUptime(hist.Statuses))
if !site.LastCheck.IsZero() {
row("Last Check", site.LastCheck.Format("15:04:05"))
}
if site.Type == "http" { if site.Type == "http" {
section("HTTP") row("Method", site.Method)
if site.Method != "" && site.Method != "GET" { row("Codes", site.AcceptedCodes)
row("Method", site.Method)
}
codes := site.AcceptedCodes
if codes == "" {
codes = "200-299"
}
row("Codes", codes)
row("SSL", fmtSSL(site)) row("SSL", fmtSSL(site))
if site.IgnoreTLS { if site.IgnoreTLS {
row("TLS Verify", dangerStyle.Render("disabled")) row("TLS Verify", dangerStyle.Render("disabled"))
} }
} }
if site.MaxRetries > 0 || site.Regions != "" || site.Description != "" { if site.MaxRetries > 0 {
section("CONFIG") row("Retries", fmtRetries(site))
if site.MaxRetries > 0 { }
row("Retries", fmtRetries(site)) if site.Regions != "" {
} row("Regions", site.Regions)
if site.Regions != "" { }
row("Regions", site.Regions) if site.Description != "" {
} row("Description", site.Description)
if site.Description != "" { }
row("Description", site.Description) if !site.LastCheck.IsZero() {
} row("Last Check", site.LastCheck.Format("15:04:05"))
} }
probeResults := m.engine.GetProbeResults(site.ID) probeResults := m.engine.GetProbeResults(site.ID)
@@ -915,30 +787,7 @@ func (m Model) viewDetailPanel() string {
} }
latency := time.Duration(result.LatencyNs).Milliseconds() latency := time.Duration(result.LatencyNs).Milliseconds()
ago := time.Since(result.CheckedAt).Truncate(time.Second) ago := time.Since(result.CheckedAt).Truncate(time.Second)
line := fmt.Sprintf(" %-14s %s %dms %s ago", nodeID, status, latency, ago) fmt.Fprintf(&b, " %-14s %s %dms %s ago\n", nodeID, status, latency, ago)
if !result.IsUp && result.ErrorReason != "" {
line += " " + dangerStyle.Render(limitStr(result.ErrorReason, 30))
}
b.WriteString(line + "\n")
}
}
stateChanges := m.engine.GetStateChanges(site.ID, 5)
if len(stateChanges) > 0 {
b.WriteString("\n" + subtleStyle.Render(" STATE CHANGES") + "\n")
for _, sc := range stateChanges {
ago := fmtDuration(time.Since(sc.ChangedAt))
arrow := subtleStyle.Render(sc.FromStatus) + " → "
if sc.ToStatus == "UP" {
arrow += specialStyle.Render(sc.ToStatus)
} else {
arrow += dangerStyle.Render(sc.ToStatus)
}
line := fmt.Sprintf(" %s %s", arrow, subtleStyle.Render(ago+" ago"))
if sc.ErrorReason != "" && sc.ToStatus != "UP" {
line += " " + dangerStyle.Render(limitStr(sc.ErrorReason, 40))
}
b.WriteString(line + "\n")
} }
} }
@@ -958,7 +807,7 @@ func (m Model) viewDetailPanel() string {
up, len(hist.Statuses)) up, len(hist.Statuses))
} }
} else { } else {
b.WriteString(" " + latencySparkline(hist.Latencies, hist.Statuses, sparkWidth)) b.WriteString(" " + latencySparkline(hist.Latencies, sparkWidth))
if len(hist.Latencies) > 0 { if len(hist.Latencies) > 0 {
minL, maxL := hist.Latencies[0], hist.Latencies[0] minL, maxL := hist.Latencies[0], hist.Latencies[0]
var total time.Duration var total time.Duration
+3 -14
View File
@@ -32,19 +32,8 @@ func (m Model) viewUsersTab() string {
return "\n No users configured. Press [n] to add one." return "\n No users configured. Press [n] to add one."
} }
var headers []string
var widths []int
if m.isWide() {
headers = []string{"#", "USERNAME", "ROLE", "PUBLIC KEY"}
widths = []int{4, 18, 10, 50}
} else {
headers = []string{"#", "USER", "ROLE", "KEY"}
widths = []int{4, 14, 8, 30}
}
userW := widths[1]
return m.renderTable( return m.renderTable(
headers, []string{"#", "USERNAME", "ROLE", "PUBLIC KEY"},
len(m.users), len(m.users),
func(start, end int) [][]string { func(start, end int) [][]string {
var rows [][]string var rows [][]string
@@ -52,14 +41,14 @@ func (m Model) viewUsersTab() string {
u := m.users[i] u := m.users[i]
rows = append(rows, []string{ rows = append(rows, []string{
fmt.Sprintf("%d", i+1), fmt.Sprintf("%d", i+1),
m.zones.Mark(fmt.Sprintf("user-%d", i), limitStr(u.Username, userW-2)), m.zones.Mark(fmt.Sprintf("user-%d", i), limitStr(u.Username, 15)),
fmtRole(u.Role), fmtRole(u.Role),
fmtKey(u.PublicKey), fmtKey(u.PublicKey),
}) })
} }
return rows return rows
}, },
widths, nil, nil, nil,
) )
} }
+4 -23
View File
@@ -15,12 +15,6 @@ var (
type StyleOverride func(row, col int) *lipgloss.Style type StyleOverride func(row, col int) *lipgloss.Style
const wideBreakpoint = 120
func (m Model) isWide() bool {
return m.termWidth >= wideBreakpoint
}
func (m Model) renderTable(headers []string, items int, buildRows func(start, end int) [][]string, colWidths []int, styleOverride StyleOverride) string { func (m Model) renderTable(headers []string, items int, buildRows func(start, end int) [][]string, colWidths []int, styleOverride StyleOverride) string {
if items == 0 { if items == 0 {
return "" return ""
@@ -34,16 +28,7 @@ func (m Model) renderTable(headers []string, items int, buildRows func(start, en
selectedVisual := m.cursor - m.tableOffset selectedVisual := m.cursor - m.tableOffset
rows := buildRows(m.tableOffset, end) rows := buildRows(m.tableOffset, end)
colTotal := 0 tableWidth := m.termWidth - chromePadH - 2
for _, w := range colWidths {
colTotal += w
}
borderOverhead := 2 + len(colWidths) - 1
tableWidth := colTotal + borderOverhead
maxWidth := m.termWidth - chromePadH - 2
if tableWidth > maxWidth {
tableWidth = maxWidth
}
if tableWidth < 40 { if tableWidth < 40 {
tableWidth = 40 tableWidth = 40
} }
@@ -56,11 +41,7 @@ func (m Model) renderTable(headers []string, items int, buildRows func(start, en
Rows(rows...). Rows(rows...).
StyleFunc(func(row, col int) lipgloss.Style { StyleFunc(func(row, col int) lipgloss.Style {
if row == table.HeaderRow { if row == table.HeaderRow {
h := tableHeaderStyle return tableHeaderStyle
if col < len(colWidths) && colWidths[col] > 0 {
h = h.Width(colWidths[col]).MaxWidth(colWidths[col])
}
return h
} }
isSelected := row == selectedVisual isSelected := row == selectedVisual
if styleOverride != nil { if styleOverride != nil {
@@ -70,7 +51,7 @@ func (m Model) renderTable(headers []string, items int, buildRows func(start, en
style = tableSelectedStyle.Foreground(s.GetForeground()) style = tableSelectedStyle.Foreground(s.GetForeground())
} }
if col < len(colWidths) && colWidths[col] > 0 { if col < len(colWidths) && colWidths[col] > 0 {
style = style.Width(colWidths[col]).MaxWidth(colWidths[col]) style = style.Width(colWidths[col])
} }
return style return style
} }
@@ -83,7 +64,7 @@ func (m Model) renderTable(headers []string, items int, buildRows func(start, en
base = tableSelectedStyle base = tableSelectedStyle
} }
if col < len(colWidths) && colWidths[col] > 0 { if col < len(colWidths) && colWidths[col] > 0 {
base = base.Width(colWidths[col]).MaxWidth(colWidths[col]) base = base.Width(colWidths[col])
} }
return base return base
}) })
+7 -56
View File
@@ -68,7 +68,6 @@ const (
stateLogs stateLogs
stateUsers stateUsers
stateDetail stateDetail
stateAlertDetail
stateFormSite stateFormSite
stateFormAlert stateFormAlert
stateFormUser stateFormUser
@@ -93,10 +92,9 @@ type Model struct {
userFormData *userFormData userFormData *userFormData
maintFormData *maintFormData maintFormData *maintFormData
logViewport viewport.Model logViewport viewport.Model
logFilterImportant bool isAdmin bool
isAdmin bool zones *zone.Manager
zones *zone.Manager
deleteID int deleteID int
deleteName string deleteName string
@@ -385,14 +383,6 @@ func (m Model) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
return m, tea.Quit return m, tea.Quit
} }
return m, nil return m, nil
case stateAlertDetail:
switch msg.String() {
case "i", "esc":
m.state = stateDashboard
case "q":
return m, tea.Quit
}
return m, nil
case stateDashboard, stateLogs, stateUsers: case stateDashboard, stateLogs, stateUsers:
switch msg.String() { switch msg.String() {
case "q": case "q":
@@ -402,11 +392,6 @@ func (m Model) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
m.filterMode = true m.filterMode = true
return m, nil return m, nil
} }
case "f":
if m.state == stateLogs {
m.logFilterImportant = !m.logFilterImportant
return m, nil
}
case "tab": case "tab":
m.switchTab(m.currentTab + 1) m.switchTab(m.currentTab + 1)
case "pgup", "pgdown": case "pgup", "pgdown":
@@ -478,16 +463,6 @@ func (m Model) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
m.state = stateFormUser m.state = stateFormUser
return m, m.initUserHuhForm() return m, m.initUserHuhForm()
} }
case "t":
if m.currentTab == 1 && len(m.alerts) > 0 {
a := m.alerts[m.cursor]
go func() {
if err := m.engine.TestAlert(a.ID); err != nil {
m.engine.AddLog(fmt.Sprintf("Test alert failed (%s): %v", a.Name, err))
}
}()
return m, nil
}
case " ": case " ":
if m.currentTab == 0 && len(m.sites) > 0 && m.sites[m.cursor].Type == "group" { if m.currentTab == 0 && len(m.sites) > 0 && m.sites[m.cursor].Type == "group" {
gid := m.sites[m.cursor].ID gid := m.sites[m.cursor].ID
@@ -506,8 +481,6 @@ func (m Model) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
case "i": case "i":
if m.currentTab == 0 && len(m.sites) > 0 { if m.currentTab == 0 && len(m.sites) > 0 {
m.state = stateDetail m.state = stateDetail
} else if m.currentTab == 1 && len(m.alerts) > 0 {
m.state = stateAlertDetail
} }
case "x": case "x":
if m.currentTab == 4 && len(m.maintenanceWindows) > 0 { if m.currentTab == 4 && len(m.maintenanceWindows) > 0 {
@@ -829,8 +802,6 @@ func (m Model) View() string {
return "" return ""
case stateDetail: case stateDetail:
return m.viewDetailPanel() return m.viewDetailPanel()
case stateAlertDetail:
return m.viewAlertDetailPanel()
default: default:
return m.zones.Scan(m.viewDashboard()) return m.zones.Scan(m.viewDashboard())
} }
@@ -840,20 +811,13 @@ func (m Model) viewDashboard() string {
allSites := m.engine.GetAllSites() allSites := m.engine.GetAllSites()
totalMonitors := 0 totalMonitors := 0
downCount := 0 downCount := 0
lateCount := 0
for _, s := range allSites { for _, s := range allSites {
if s.Type == "group" { if s.Type == "group" {
continue continue
} }
totalMonitors++ totalMonitors++
if s.Paused || m.isMonitorInMaintenance(s.ID) { if !s.Paused && !m.isMonitorInMaintenance(s.ID) && (s.Status == "DOWN" || s.Status == "SSL EXP") {
continue
}
switch s.Status {
case "DOWN", "SSL EXP":
downCount++ downCount++
case "LATE":
lateCount++
} }
} }
offlineNodes := 0 offlineNodes := 0
@@ -866,8 +830,6 @@ func (m Model) viewDashboard() string {
var sitesLabel string var sitesLabel string
if downCount > 0 { if downCount > 0 {
sitesLabel = fmt.Sprintf("Sites (%d↓)", downCount) sitesLabel = fmt.Sprintf("Sites (%d↓)", downCount)
} else if lateCount > 0 {
sitesLabel = fmt.Sprintf("Sites (%d⚠)", lateCount)
} else if totalMonitors > 0 { } else if totalMonitors > 0 {
sitesLabel = fmt.Sprintf("Sites (%d)", totalMonitors) sitesLabel = fmt.Sprintf("Sites (%d)", totalMonitors)
} else { } else {
@@ -933,19 +895,14 @@ func (m Model) viewDashboard() string {
} }
} }
upCount := totalMonitors - downCount - lateCount upCount := totalMonitors - downCount
var upStr string var upStr string
if downCount > 0 { if downCount > 0 {
upStr = dangerStyle.Render(fmt.Sprintf("%d/%d UP", upCount, totalMonitors)) upStr = dangerStyle.Render(fmt.Sprintf("%d/%d UP", upCount, totalMonitors))
} else if lateCount > 0 {
upStr = warnStyle.Render(fmt.Sprintf("%d/%d UP", upCount, totalMonitors))
} else { } else {
upStr = specialStyle.Render(fmt.Sprintf("%d/%d UP", upCount, totalMonitors)) upStr = specialStyle.Render(fmt.Sprintf("%d/%d UP", upCount, totalMonitors))
} }
statusParts := []string{upStr} statusParts := []string{upStr}
if lateCount > 0 {
statusParts = append(statusParts, warnStyle.Render(fmt.Sprintf("%d LATE", lateCount)))
}
if len(m.nodes) > 0 { if len(m.nodes) > 0 {
online := 0 online := 0
for _, n := range m.nodes { for _, n := range m.nodes {
@@ -966,10 +923,6 @@ func (m Model) viewDashboard() string {
switch m.currentTab { switch m.currentTab {
case 0: case 0:
keys = "[/]Filter [n]New [e]Edit [i]Info [d]Del [p]Pause [T]Theme [Tab]Switch [q]Quit" keys = "[/]Filter [n]New [e]Edit [i]Info [d]Del [p]Pause [T]Theme [Tab]Switch [q]Quit"
case 1:
keys = "[n]New [e]Edit [i]Info [d]Del [t]Test [T]Theme [Tab]Switch [q]Quit"
case 2:
keys = "[f]Filter [T]Theme [Tab]Switch [q]Quit"
case 4: case 4:
keys = "[n]New [x]End [d]Del [T]Theme [Tab]Switch [q]Quit" keys = "[n]New [x]End [d]Del [T]Theme [Tab]Switch [q]Quit"
case 5: case 5:
@@ -996,12 +949,10 @@ func siteOrder(s models.Site) int {
switch s.Status { switch s.Status {
case "DOWN", "SSL EXP": case "DOWN", "SSL EXP":
return 0 return 0
case "LATE":
return 1
case "PENDING": case "PENDING":
return 3
default:
return 2 return 2
default:
return 1
} }
} }
-274
View File
@@ -1,274 +0,0 @@
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()
}
-54
View File
@@ -1,54 +0,0 @@
Set Shell "bash"
Set Width 1400
Set Height 800
Set FontSize 14
Set Padding 20
Set Framerate 15
Set TypingSpeed 50ms
Hide
Type "bash vhs/setup.sh /tmp/uptop-vhs.db"
Enter
Sleep 45s
Show
Sleep 5s
# Sites tab — hero shot with mixed monitor states
Screenshot vhs/screenshots/monitors.png
Sleep 1s
# Navigate to Nextcloud (row 6: group + 3 children + Auth Portal)
Down
Sleep 200ms
Down
Sleep 200ms
Down
Sleep 200ms
Down
Sleep 200ms
Down
Sleep 200ms
Type "i"
Sleep 3s
Screenshot vhs/screenshots/detail.png
Sleep 1s
# Close detail
Escape
Sleep 1s
# Tab to Alerts
Tab
Sleep 2s
Screenshot vhs/screenshots/alerts.png
Sleep 1s
# Tab to Logs
Tab
Sleep 2s
Screenshot vhs/screenshots/logs.png
Sleep 1s
# Quit
Type "q"
Sleep 1s
Binary file not shown.

Before

Width:  |  Height:  |  Size: 84 KiB

After

Width:  |  Height:  |  Size: 60 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 80 KiB

After

Width:  |  Height:  |  Size: 53 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 160 KiB

After

Width:  |  Height:  |  Size: 5.1 KiB

Binary file not shown.

Before

Width:  |  Height:  |  Size: 219 KiB

After

Width:  |  Height:  |  Size: 172 KiB

-141
View File
@@ -1,141 +0,0 @@
alerts:
- name: Discord Homelab
type: discord
settings:
url: https://discord.com/api/webhooks/1234567890/demo-token
- name: Ntfy Alerts
type: webhook
settings:
url: https://ntfy.example.com/homelab-alerts
- name: Email Oncall
type: email
settings:
host: smtp.example.com
port: "587"
user: alerts@example.com
pass: "••••••••"
from: alerts@example.com
to: oncall@example.com
- name: Slack Ops
type: slack
settings:
url: https://hooks.slack.com/services/T00000/B00000/demo-token
monitors:
# HTTP — homelab services
- name: Nextcloud
type: http
url: https://example.com
interval: 30
alert: Discord Homelab
check_ssl: true
expiry_threshold: 14
max_retries: 2
- name: Jellyfin
type: http
url: https://example.com
interval: 30
alert: Discord Homelab
max_retries: 2
- name: Home Assistant
type: http
url: https://example.com
interval: 30
alert: Discord Homelab
max_retries: 3
- name: Gitea
type: http
url: https://example.com
interval: 60
alert: Discord Homelab
check_ssl: true
expiry_threshold: 14
max_retries: 2
- name: Traefik Dashboard
type: http
url: https://example.com
interval: 60
alert: Discord Homelab
max_retries: 1
- name: Vaultwarden
type: http
url: https://example.com
interval: 30
alert: Discord Homelab
check_ssl: true
expiry_threshold: 14
max_retries: 3
- name: Personal Blog
type: http
url: https://example.com
interval: 120
alert: Discord Homelab
check_ssl: true
expiry_threshold: 14
max_retries: 2
- name: Immich
type: http
url: https://example.com
interval: 60
alert: Discord Homelab
check_ssl: true
expiry_threshold: 7
max_retries: 3
# HTTP — deliberate failure
- name: Auth Portal
type: http
url: http://localhost:1
interval: 30
alert: Discord Homelab
max_retries: 2
# Push — cron jobs
- name: Nightly Backup
type: push
interval: 300
alert: Discord Homelab
- name: Cert Renewal
type: push
interval: 300
alert: Discord Homelab
# Infrastructure group
- name: Infrastructure
type: group
alert: Discord Homelab
monitors:
- name: Edge Router
type: ping
hostname: 8.8.8.8
interval: 30
alert: Discord Homelab
timeout: 5
- name: Postgres
type: port
hostname: localhost
port: 18099
interval: 60
alert: Discord Homelab
timeout: 5
- name: DNS Primary
type: dns
hostname: google.com
dns_server: 8.8.8.8
dns_resolve_type: A
interval: 60
alert: Discord Homelab
timeout: 5
-27
View File
@@ -1,27 +0,0 @@
#!/bin/bash
# VHS screenshot setup: seed monitors, backfill history, start server.
set -e
DB="${1:?usage: setup.sh <db-path>}"
rm -f "$DB" "$DB-shm" "$DB-wal"
echo "==> Seeding monitors and alerts..."
UPTOP_DB_DSN="$DB" ./uptop apply -f vhs/seed.yaml 2>&1
echo "==> Backfilling check history..."
BACKFILL_OUT=$(go run ./vhs/backfill/ "$DB")
echo "$BACKFILL_OUT"
PUSH_TOKEN=$(echo "$BACKFILL_OUT" | grep '^PUSH_TOKEN=' | cut -d= -f2)
if [ -n "$PUSH_TOKEN" ]; then
echo "==> Sending push heartbeat in 15s (background)..."
(sleep 15 && curl -s "http://localhost:18099/api/push" -H "Authorization: Bearer $PUSH_TOKEN" > /dev/null 2>&1) &
fi
echo "==> Starting uptop server..."
exec env \
UPTOP_DB_DSN="$DB" \
UPTOP_PORT=23299 \
UPTOP_HTTP_PORT=18099 \
UPTOP_ALLOW_PRIVATE_TARGETS=true \
./uptop serve 2>/dev/null