mirror of
https://github.com/Rain-kl/OpenFlare.git
synced 2026-10-04 23:16:37 +08:00
feat(obs): 访问日志 SSOT 与 edge_health,去掉协议兼容层
Agent 仅上报 host_metrics/edge_health/access_logs;业务流量与 UV 由 Server 侧访问日志聚合。新增 of_node_edge_health 与 of_access_log_hourly, 删除 request_reports/openresty 吞吐路径;API 不再暴露 traffic_reports 与 openresty_rx|tx。心跳/离线默认阈值与回填迁移一并入库。
This commit is contained in:
@@ -16,6 +16,7 @@ const (
|
||||
maxAccessLogPageSize = 200
|
||||
defaultAccessLogSortBy = "logged_at"
|
||||
defaultAccessLogSortOrder = "desc"
|
||||
accessLogSortOrderAsc = "asc"
|
||||
defaultAccessLogFoldMinute = 3
|
||||
defaultIPTrendHours = 24
|
||||
defaultIPTrendBucketMinute = 30
|
||||
@@ -605,8 +606,8 @@ func normalizeAccessLogSortBy(sortBy string) string {
|
||||
}
|
||||
|
||||
func normalizeAccessLogSortOrder(sortOrder string) string {
|
||||
if strings.EqualFold(strings.TrimSpace(sortOrder), "asc") {
|
||||
return "asc"
|
||||
if strings.EqualFold(strings.TrimSpace(sortOrder), accessLogSortOrderAsc) {
|
||||
return accessLogSortOrderAsc
|
||||
}
|
||||
return defaultAccessLogSortOrder
|
||||
}
|
||||
|
||||
@@ -5,7 +5,6 @@ package observability
|
||||
|
||||
import (
|
||||
"context"
|
||||
"encoding/json"
|
||||
"sort"
|
||||
"strings"
|
||||
"time"
|
||||
@@ -22,6 +21,7 @@ const (
|
||||
healthSeverityCritical = "critical"
|
||||
healthSeverityWarning = "warning"
|
||||
percentageMultiplier = 100
|
||||
sortOrderAsc = "asc"
|
||||
)
|
||||
|
||||
// DistributionItem is a key/value distribution entry.
|
||||
@@ -37,9 +37,9 @@ type TrafficDistributions struct {
|
||||
SourceCountries []DistributionItem `json:"source_countries"`
|
||||
}
|
||||
|
||||
const metricSnapshotOpenrestyMatchWindow = 2 * time.Minute
|
||||
const metricSnapshotEdgeHealthMatchWindow = 2 * time.Minute
|
||||
|
||||
// NodeMetricSnapshotView is a metric snapshot enriched with OpenResty observations.
|
||||
// NodeMetricSnapshotView is a metric snapshot enriched with edge health connections.
|
||||
type NodeMetricSnapshotView struct {
|
||||
ID uint `json:"id,omitempty"`
|
||||
NodeID string `json:"node_id,omitempty"`
|
||||
@@ -53,8 +53,6 @@ type NodeMetricSnapshotView struct {
|
||||
DiskWriteBytes int64 `json:"disk_write_bytes"`
|
||||
NetworkRxBytes int64 `json:"network_rx_bytes"`
|
||||
NetworkTxBytes int64 `json:"network_tx_bytes"`
|
||||
OpenrestyRxBytes int64 `json:"openresty_rx_bytes"`
|
||||
OpenrestyTxBytes int64 `json:"openresty_tx_bytes"`
|
||||
OpenrestyConnections int64 `json:"openresty_connections"`
|
||||
}
|
||||
|
||||
@@ -98,13 +96,15 @@ type CapacityTrendPoint struct {
|
||||
}
|
||||
|
||||
// NetworkTrendPoint is a network trend bucket.
|
||||
// Host network_* is L3 (宿主机网卡).
|
||||
// bytes_received/provided are L1 business bytes from access logs.
|
||||
type NetworkTrendPoint struct {
|
||||
BucketStartedAt time.Time `json:"bucket_started_at"`
|
||||
NetworkRxBytes int64 `json:"network_rx_bytes"`
|
||||
NetworkTxBytes int64 `json:"network_tx_bytes"`
|
||||
OpenrestyRxBytes int64 `json:"openresty_rx_bytes"`
|
||||
OpenrestyTxBytes int64 `json:"openresty_tx_bytes"`
|
||||
ReportedNodes int `json:"reported_nodes"`
|
||||
BucketStartedAt time.Time `json:"bucket_started_at"`
|
||||
NetworkRxBytes int64 `json:"network_rx_bytes"`
|
||||
NetworkTxBytes int64 `json:"network_tx_bytes"`
|
||||
BytesReceived int64 `json:"bytes_received"` // sum(request_length)
|
||||
BytesProvided int64 `json:"bytes_provided"` // sum(bytes_sent)
|
||||
ReportedNodes int `json:"reported_nodes"`
|
||||
}
|
||||
|
||||
// DiskIOTrendPoint is a disk IO trend bucket.
|
||||
@@ -141,30 +141,39 @@ type networkCounterState struct {
|
||||
seen bool
|
||||
}
|
||||
|
||||
func buildTrafficWindowSummary(report *model.OpenFlareRequestReport) *TrafficWindowSummary {
|
||||
if report == nil {
|
||||
func buildTrafficWindowSummaryFromAccessLogs(
|
||||
ctx context.Context,
|
||||
nodeID string,
|
||||
since, until time.Time,
|
||||
) *TrafficWindowSummary {
|
||||
row, err := model.TrafficSummaryOpenFlareAccessLogs(ctx, model.OpenFlareAccessLogQuery{
|
||||
NodeID: nodeID,
|
||||
Since: since,
|
||||
Until: until,
|
||||
})
|
||||
if err != nil || row.RequestCount <= 0 {
|
||||
return nil
|
||||
}
|
||||
summary := TrafficWindowSummary{
|
||||
WindowStartedAt: report.WindowStartedAt,
|
||||
WindowEndedAt: report.WindowEndedAt,
|
||||
RequestCount: report.RequestCount,
|
||||
UniqueVisitorCount: report.UniqueVisitorCount,
|
||||
ErrorCount: report.ErrorCount,
|
||||
summary := &TrafficWindowSummary{
|
||||
WindowStartedAt: since.UTC(),
|
||||
WindowEndedAt: until.UTC(),
|
||||
RequestCount: row.RequestCount,
|
||||
UniqueVisitorCount: row.UniqueIPCount,
|
||||
ErrorCount: row.ErrorCount,
|
||||
}
|
||||
if duration := report.WindowEndedAt.Sub(report.WindowStartedAt).Seconds(); duration > 0 {
|
||||
summary.EstimatedQPS = float64(report.RequestCount) / duration
|
||||
if duration := until.Sub(since).Seconds(); duration > 0 {
|
||||
summary.EstimatedQPS = float64(row.RequestCount) / duration
|
||||
}
|
||||
if report.RequestCount > 0 {
|
||||
summary.ErrorRatePercent = (float64(report.ErrorCount) / float64(report.RequestCount)) * 100
|
||||
if row.RequestCount > 0 {
|
||||
summary.ErrorRatePercent = (float64(row.ErrorCount) / float64(row.RequestCount)) * 100
|
||||
}
|
||||
return &summary
|
||||
return summary
|
||||
}
|
||||
|
||||
// BuildMetricSnapshotViews merges metric snapshots with OpenResty observations for API responses.
|
||||
// BuildMetricSnapshotViews merges metric snapshots with edge health connections for API responses.
|
||||
func BuildMetricSnapshotViews(
|
||||
snapshots []*model.OpenFlareMetricSnapshot,
|
||||
openrestyObs []*model.OpenFlareNodeObservationOpenresty,
|
||||
edgeHealth []*model.OpenFlareEdgeHealth,
|
||||
) []*NodeMetricSnapshotView {
|
||||
if len(snapshots) == 0 {
|
||||
return []*NodeMetricSnapshotView{}
|
||||
@@ -188,40 +197,49 @@ func BuildMetricSnapshotViews(
|
||||
NetworkRxBytes: snapshot.NetworkRxBytes,
|
||||
NetworkTxBytes: snapshot.NetworkTxBytes,
|
||||
}
|
||||
if matched := matchOpenrestyObservation(snapshot.CapturedAt, openrestyObs); matched != nil {
|
||||
view.OpenrestyRxBytes = matched.OpenrestyRxBytes
|
||||
view.OpenrestyTxBytes = matched.OpenrestyTxBytes
|
||||
view.OpenrestyConnections = matched.OpenrestyConnections
|
||||
if matched := matchEdgeHealth(snapshot.CapturedAt, edgeHealth); matched != nil {
|
||||
view.OpenrestyConnections = matched.Connections
|
||||
}
|
||||
views = append(views, view)
|
||||
}
|
||||
return views
|
||||
}
|
||||
|
||||
// BuildTrafficDistributions aggregates traffic distribution charts.
|
||||
func BuildTrafficDistributions(
|
||||
reports []*model.OpenFlareRequestReport,
|
||||
accessLogRegions []*model.OpenFlareAccessLogRegionCount,
|
||||
// BuildTrafficDistributionsFromAccessLogs builds distributions from access logs (L1).
|
||||
func BuildTrafficDistributionsFromAccessLogs(
|
||||
ctx context.Context,
|
||||
since, until time.Time,
|
||||
limit int,
|
||||
accessLogRegions []*model.OpenFlareAccessLogRegionCount,
|
||||
) TrafficDistributions {
|
||||
statusCodes := make(distributionAccumulator)
|
||||
topDomains := make(distributionAccumulator)
|
||||
reportSourceCountries := make(distributionAccumulator)
|
||||
for _, report := range reports {
|
||||
mergeJSONCounts(statusCodes, report.StatusCodesJSON)
|
||||
mergeJSONCounts(topDomains, report.TopDomainsJSON)
|
||||
mergeJSONCounts(reportSourceCountries, report.SourceCountriesJSON)
|
||||
}
|
||||
sourceCountries := reportSourceCountries
|
||||
if len(accessLogRegions) > 0 {
|
||||
sourceCountries = make(distributionAccumulator, len(accessLogRegions))
|
||||
for _, item := range accessLogRegions {
|
||||
if item == nil || strings.TrimSpace(item.Region) == "" || item.Count <= 0 {
|
||||
|
||||
query := model.OpenFlareAccessLogQuery{Since: since, Until: until}
|
||||
if statusRows, err := model.ValueCountsOpenFlareAccessLogs(ctx, query, "status_code", limit); err == nil {
|
||||
for _, row := range statusRows {
|
||||
if strings.TrimSpace(row.Value) == "" || row.Count <= 0 {
|
||||
continue
|
||||
}
|
||||
sourceCountries[item.Region] = item.Count
|
||||
statusCodes[row.Value] = row.Count
|
||||
}
|
||||
}
|
||||
if hostRows, err := model.ValueCountsOpenFlareAccessLogs(ctx, query, "host", limit); err == nil {
|
||||
for _, row := range hostRows {
|
||||
if strings.TrimSpace(row.Value) == "" || row.Count <= 0 {
|
||||
continue
|
||||
}
|
||||
topDomains[row.Value] = row.Count
|
||||
}
|
||||
}
|
||||
|
||||
sourceCountries := make(distributionAccumulator)
|
||||
for _, item := range accessLogRegions {
|
||||
if item == nil || strings.TrimSpace(item.Region) == "" || item.Count <= 0 {
|
||||
continue
|
||||
}
|
||||
sourceCountries[item.Region] = item.Count
|
||||
}
|
||||
return TrafficDistributions{
|
||||
StatusCodes: toDistributionItems(statusCodes, limit),
|
||||
TopDomains: toDistributionItems(topDomains, limit),
|
||||
@@ -231,7 +249,7 @@ func BuildTrafficDistributions(
|
||||
|
||||
func buildHealthSummary(
|
||||
snapshot *model.OpenFlareMetricSnapshot,
|
||||
report *model.OpenFlareRequestReport,
|
||||
traffic *TrafficWindowSummary,
|
||||
events []*model.OpenFlareHealthEvent,
|
||||
) HealthSummary {
|
||||
summary := HealthSummary{}
|
||||
@@ -258,41 +276,37 @@ func buildHealthSummary(
|
||||
storageUsage := Percentage(snapshot.StorageUsedBytes, snapshot.StorageTotalBytes)
|
||||
summary.HasCapacityRisk = snapshot.CPUUsagePercent >= 80 || memoryUsage >= 85 || storageUsage >= 85
|
||||
}
|
||||
if report != nil && report.RequestCount >= 100 {
|
||||
summary.HasTrafficRisk = (float64(report.ErrorCount) / float64(report.RequestCount)) >= 0.05
|
||||
if traffic != nil && traffic.RequestCount >= 100 {
|
||||
summary.HasTrafficRisk = (float64(traffic.ErrorCount) / float64(traffic.RequestCount)) >= 0.05
|
||||
}
|
||||
summary.HasRuntimeRisk = summary.ActiveAlerts > 0 || summary.HasCapacityRisk || summary.HasTrafficRisk
|
||||
return summary
|
||||
}
|
||||
|
||||
// BuildNodeTrends builds 24h trend series, preferring ClickHouse hourly aggregates
|
||||
// over limited raw snapshot windows so capacity/network/disk charts stay complete.
|
||||
// BuildNodeTrends builds 24h trend series.
|
||||
// Business traffic (requests/errors/UV and provided/received bytes) comes from access logs.
|
||||
// Host capacity/disk/network come from metric snapshots (hourly when available).
|
||||
func BuildNodeTrends(
|
||||
ctx context.Context,
|
||||
now time.Time,
|
||||
nodeID string,
|
||||
snapshots []*model.OpenFlareMetricSnapshot,
|
||||
openrestyObs []*model.OpenFlareNodeObservationOpenresty,
|
||||
reports []*model.OpenFlareRequestReport,
|
||||
) NodeTrends {
|
||||
trendSince := now.Add(-24 * time.Hour)
|
||||
trafficTrend := BuildTrafficTrendPoints(now, reports)
|
||||
if trafficHourly, err := model.ListOpenFlareTrafficHourlySince(ctx, nodeID, trendSince); err == nil && len(trafficHourly) > 0 {
|
||||
trafficTrend = BuildTrafficTrendPointsFromHourly(now, trafficHourly)
|
||||
}
|
||||
|
||||
trafficTrend := BuildTrafficTrendPointsFromAccessLogs(ctx, now, nodeID, trendSince)
|
||||
capacityTrend := BuildCapacityTrendPoints(now, snapshots)
|
||||
networkTrend := BuildNetworkTrendPoints(now, snapshots, openrestyObs)
|
||||
networkTrend := BuildNetworkTrendPoints(now, snapshots)
|
||||
// Overlay L1 business bytes onto network points.
|
||||
applyAccessLogBytesToNetworkTrend(ctx, now, nodeID, trendSince, networkTrend)
|
||||
diskIOTrend := BuildDiskIOTrendPoints(now, snapshots)
|
||||
|
||||
metricHourly, metricErr := model.ListOpenFlareMetricHourlySince(ctx, nodeID, trendSince)
|
||||
if metricErr == nil && len(metricHourly) > 0 {
|
||||
capacityTrend = BuildCapacityTrendPointsFromHourly(now, metricHourly)
|
||||
diskIOTrend = BuildDiskIOTrendPointsFromHourly(now, metricHourly)
|
||||
}
|
||||
openrestyHourly, openrestyErr := model.ListOpenFlareOpenrestyHourlySince(ctx, nodeID, trendSince)
|
||||
if metricErr == nil && openrestyErr == nil && (len(metricHourly) > 0 || len(openrestyHourly) > 0) {
|
||||
networkTrend = BuildNetworkTrendPointsFromHourly(now, metricHourly, openrestyHourly)
|
||||
networkTrend = BuildNetworkTrendPointsFromHourly(now, metricHourly)
|
||||
applyAccessLogBytesToNetworkTrend(ctx, now, nodeID, trendSince, networkTrend)
|
||||
}
|
||||
|
||||
return NodeTrends{
|
||||
@@ -303,7 +317,131 @@ func BuildNodeTrends(
|
||||
}
|
||||
}
|
||||
|
||||
// BuildTrafficTrendPointsFromAccessLogs builds 24h request/error buckets from access logs.
|
||||
// Prefers of_access_log_hourly when available; falls back to raw bucket aggregates.
|
||||
// UniqueVisitorCount on hourly path is 0 (use TrafficSummary for exact UV).
|
||||
func BuildTrafficTrendPointsFromAccessLogs(ctx context.Context, now time.Time, nodeID string, since time.Time) []TrafficTrendPoint {
|
||||
start := trendWindowStart(now)
|
||||
points := make([]TrafficTrendPoint, observabilityTrendBuckets)
|
||||
for index := range points {
|
||||
points[index].BucketStartedAt = start.Add(time.Duration(index) * time.Hour)
|
||||
}
|
||||
|
||||
if hourly, err := model.ListOpenFlareTrafficHourlySince(ctx, nodeID, since); err == nil && len(hourly) > 0 {
|
||||
for _, row := range hourly {
|
||||
if row == nil {
|
||||
continue
|
||||
}
|
||||
index, ok := trendBucketIndex(row.Hour, start)
|
||||
if !ok {
|
||||
continue
|
||||
}
|
||||
points[index].RequestCount += row.RequestCount
|
||||
points[index].ErrorCount += row.ErrorCount
|
||||
// UniqueVisitorCount intentionally not summed from hourly rollup (always 0 / overcounts).
|
||||
}
|
||||
return points
|
||||
}
|
||||
|
||||
buckets, err := model.ListOpenFlareAccessLogBuckets(ctx, model.OpenFlareAccessLogBucketQuery{
|
||||
NodeID: nodeID,
|
||||
Since: since,
|
||||
Until: now,
|
||||
FoldMinutes: 60,
|
||||
SortBy: "logged_at",
|
||||
SortOrder: sortOrderAsc,
|
||||
})
|
||||
if err != nil || len(buckets) == 0 {
|
||||
return points
|
||||
}
|
||||
byEpoch := make(map[int64]*model.OpenFlareAccessLogBucketRow, len(buckets))
|
||||
for _, row := range buckets {
|
||||
if row == nil {
|
||||
continue
|
||||
}
|
||||
byEpoch[row.BucketEpoch] = row
|
||||
}
|
||||
for index := range points {
|
||||
epoch := points[index].BucketStartedAt.Unix()
|
||||
if row, ok := byEpoch[epoch]; ok {
|
||||
points[index].RequestCount = row.RequestCount
|
||||
points[index].ErrorCount = row.ServerErrorCount
|
||||
points[index].UniqueVisitorCount = row.UniqueIPCount
|
||||
}
|
||||
}
|
||||
return points
|
||||
}
|
||||
|
||||
func applyAccessLogBytesToNetworkTrend(ctx context.Context, now time.Time, nodeID string, since time.Time, points []NetworkTrendPoint) {
|
||||
if len(points) == 0 {
|
||||
return
|
||||
}
|
||||
// Prefer of_access_log_hourly (summed across hosts).
|
||||
if hourly, err := analyticsListAccessLogHourlyBytes(ctx, nodeID, since); err == nil && len(hourly) > 0 {
|
||||
for hourUnix, totals := range hourly {
|
||||
for index := range points {
|
||||
if points[index].BucketStartedAt.Unix() == hourUnix {
|
||||
points[index].BytesProvided = totals.provided
|
||||
points[index].BytesReceived = totals.received
|
||||
}
|
||||
}
|
||||
}
|
||||
return
|
||||
}
|
||||
|
||||
buckets, err := model.ListOpenFlareAccessLogBuckets(ctx, model.OpenFlareAccessLogBucketQuery{
|
||||
NodeID: nodeID,
|
||||
Since: since,
|
||||
Until: now,
|
||||
FoldMinutes: 60,
|
||||
SortBy: "logged_at",
|
||||
SortOrder: sortOrderAsc,
|
||||
})
|
||||
if err != nil || len(buckets) == 0 {
|
||||
return
|
||||
}
|
||||
byEpoch := make(map[int64]*model.OpenFlareAccessLogBucketRow, len(buckets))
|
||||
for _, row := range buckets {
|
||||
if row == nil {
|
||||
continue
|
||||
}
|
||||
byEpoch[row.BucketEpoch] = row
|
||||
}
|
||||
for index := range points {
|
||||
epoch := points[index].BucketStartedAt.Unix()
|
||||
if row, ok := byEpoch[epoch]; ok {
|
||||
points[index].BytesProvided = row.BytesSent
|
||||
points[index].BytesReceived = row.RequestLength
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
type accessLogHourBytes struct {
|
||||
provided int64
|
||||
received int64
|
||||
}
|
||||
|
||||
func analyticsListAccessLogHourlyBytes(ctx context.Context, nodeID string, since time.Time) (map[int64]accessLogHourBytes, error) {
|
||||
rows, err := model.ListOpenFlareAccessLogHourlySince(ctx, nodeID, since)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
out := make(map[int64]accessLogHourBytes)
|
||||
for _, row := range rows {
|
||||
if row == nil {
|
||||
continue
|
||||
}
|
||||
key := row.Hour.UTC().Truncate(time.Hour).Unix()
|
||||
cur := out[key]
|
||||
cur.provided += row.BytesSent
|
||||
cur.received += row.RequestLength
|
||||
out[key] = cur
|
||||
}
|
||||
return out, nil
|
||||
}
|
||||
|
||||
// BuildTrafficTrendPointsFromHourly builds 24h traffic trend buckets from hourly rollups.
|
||||
// UniqueVisitorCount is left at 0: hourly UV is not summed (use TrafficSummary for exact UV).
|
||||
func BuildTrafficTrendPointsFromHourly(now time.Time, hourly []*model.OpenFlareTrafficHourly) []TrafficTrendPoint {
|
||||
start := trendWindowStart(now)
|
||||
points := make([]TrafficTrendPoint, observabilityTrendBuckets)
|
||||
@@ -320,26 +458,6 @@ func BuildTrafficTrendPointsFromHourly(now time.Time, hourly []*model.OpenFlareT
|
||||
}
|
||||
points[index].RequestCount += row.RequestCount
|
||||
points[index].ErrorCount += row.ErrorCount
|
||||
points[index].UniqueVisitorCount += row.UniqueVisitorCount
|
||||
}
|
||||
return points
|
||||
}
|
||||
|
||||
// BuildTrafficTrendPoints builds 24h traffic trend buckets.
|
||||
func BuildTrafficTrendPoints(now time.Time, reports []*model.OpenFlareRequestReport) []TrafficTrendPoint {
|
||||
start := trendWindowStart(now)
|
||||
points := make([]TrafficTrendPoint, observabilityTrendBuckets)
|
||||
for index := range points {
|
||||
points[index].BucketStartedAt = start.Add(time.Duration(index) * time.Hour)
|
||||
}
|
||||
for _, report := range reports {
|
||||
index, ok := trendBucketIndex(report.WindowEndedAt, start)
|
||||
if !ok {
|
||||
continue
|
||||
}
|
||||
points[index].RequestCount += report.RequestCount
|
||||
points[index].ErrorCount += report.ErrorCount
|
||||
points[index].UniqueVisitorCount += report.UniqueVisitorCount
|
||||
}
|
||||
return points
|
||||
}
|
||||
@@ -404,12 +522,12 @@ func BuildCapacityTrendPointsFromHourly(now time.Time, hourly []*model.OpenFlare
|
||||
return points
|
||||
}
|
||||
|
||||
// BuildNetworkTrendPoints builds 24h network trend buckets.
|
||||
// Host and OpenResty counters are cumulative; values are consecutive deltas.
|
||||
// BuildNetworkTrendPoints builds 24h host-network trend buckets.
|
||||
// Host network counters must be process-lifetime cumulative values; this function
|
||||
// converts consecutive samples into deltas.
|
||||
func BuildNetworkTrendPoints(
|
||||
now time.Time,
|
||||
snapshots []*model.OpenFlareMetricSnapshot,
|
||||
openrestyObs []*model.OpenFlareNodeObservationOpenresty,
|
||||
) []NetworkTrendPoint {
|
||||
start := trendWindowStart(now)
|
||||
points := make([]NetworkTrendPoint, observabilityTrendBuckets)
|
||||
@@ -452,51 +570,17 @@ func BuildNetworkTrendPoints(
|
||||
accumulators[index].nodes[snapshot.NodeID] = struct{}{}
|
||||
}
|
||||
}
|
||||
sort.Slice(openrestyObs, func(i int, j int) bool {
|
||||
if openrestyObs[i].CapturedAt.Equal(openrestyObs[j].CapturedAt) {
|
||||
return openrestyObs[i].NodeID < openrestyObs[j].NodeID
|
||||
}
|
||||
return openrestyObs[i].CapturedAt.Before(openrestyObs[j].CapturedAt)
|
||||
})
|
||||
previousOpenrestyByNode := make(map[string]networkCounterState, len(openrestyObs))
|
||||
for _, obs := range openrestyObs {
|
||||
if obs == nil {
|
||||
continue
|
||||
}
|
||||
nodeKey := obs.NodeID
|
||||
if nodeKey == "" {
|
||||
nodeKey = unknownTrendNodeKey
|
||||
}
|
||||
previous := previousOpenrestyByNode[nodeKey]
|
||||
previousOpenrestyByNode[nodeKey] = networkCounterState{
|
||||
rx: obs.OpenrestyRxBytes,
|
||||
tx: obs.OpenrestyTxBytes,
|
||||
seen: true,
|
||||
}
|
||||
if !previous.seen {
|
||||
continue
|
||||
}
|
||||
index, ok := trendBucketIndex(obs.CapturedAt, start)
|
||||
if !ok {
|
||||
continue
|
||||
}
|
||||
points[index].OpenrestyRxBytes += nonNegativeDelta(obs.OpenrestyRxBytes, previous.rx)
|
||||
points[index].OpenrestyTxBytes += nonNegativeDelta(obs.OpenrestyTxBytes, previous.tx)
|
||||
if obs.NodeID != "" {
|
||||
accumulators[index].nodes[obs.NodeID] = struct{}{}
|
||||
}
|
||||
}
|
||||
for index := range points {
|
||||
points[index].ReportedNodes = len(accumulators[index].nodes)
|
||||
}
|
||||
return points
|
||||
}
|
||||
|
||||
// BuildNetworkTrendPointsFromHourly builds 24h network trend buckets from hourly aggregates.
|
||||
// BuildNetworkTrendPointsFromHourly builds 24h host-network trend buckets from metric hourly aggregates.
|
||||
// Business bytes (已提供/接收) are applied separately via applyAccessLogBytesToNetworkTrend.
|
||||
func BuildNetworkTrendPointsFromHourly(
|
||||
now time.Time,
|
||||
metricHourly []*model.OpenFlareMetricHourly,
|
||||
openrestyHourly []*model.OpenFlareOpenrestyHourly,
|
||||
) []NetworkTrendPoint {
|
||||
start := trendWindowStart(now)
|
||||
points := make([]NetworkTrendPoint, observabilityTrendBuckets)
|
||||
@@ -517,20 +601,6 @@ func BuildNetworkTrendPointsFromHourly(
|
||||
points[index].ReportedNodes = row.ReportedNodes
|
||||
}
|
||||
}
|
||||
for _, row := range openrestyHourly {
|
||||
if row == nil {
|
||||
continue
|
||||
}
|
||||
index, ok := trendBucketIndex(row.Hour, start)
|
||||
if !ok {
|
||||
continue
|
||||
}
|
||||
points[index].OpenrestyRxBytes += row.OpenrestyRxBytes
|
||||
points[index].OpenrestyTxBytes += row.OpenrestyTxBytes
|
||||
if row.ReportedNodes > points[index].ReportedNodes {
|
||||
points[index].ReportedNodes = row.ReportedNodes
|
||||
}
|
||||
}
|
||||
return points
|
||||
}
|
||||
|
||||
@@ -623,38 +693,25 @@ func latestMetricSnapshot(snapshots []*model.OpenFlareMetricSnapshot) *model.Ope
|
||||
return latest
|
||||
}
|
||||
|
||||
func latestTrafficReport(reports []*model.OpenFlareRequestReport) *model.OpenFlareRequestReport {
|
||||
var latest *model.OpenFlareRequestReport
|
||||
for _, report := range reports {
|
||||
if report == nil {
|
||||
continue
|
||||
}
|
||||
if latest == nil || report.WindowEndedAt.After(latest.WindowEndedAt) {
|
||||
latest = report
|
||||
}
|
||||
}
|
||||
return latest
|
||||
}
|
||||
|
||||
func matchOpenrestyObservation(
|
||||
func matchEdgeHealth(
|
||||
capturedAt time.Time,
|
||||
observations []*model.OpenFlareNodeObservationOpenresty,
|
||||
) *model.OpenFlareNodeObservationOpenresty {
|
||||
var matched *model.OpenFlareNodeObservationOpenresty
|
||||
bestDelta := metricSnapshotOpenrestyMatchWindow + time.Second
|
||||
for _, observation := range observations {
|
||||
if observation == nil {
|
||||
health []*model.OpenFlareEdgeHealth,
|
||||
) *model.OpenFlareEdgeHealth {
|
||||
var matched *model.OpenFlareEdgeHealth
|
||||
bestDelta := metricSnapshotEdgeHealthMatchWindow + time.Second
|
||||
for _, row := range health {
|
||||
if row == nil {
|
||||
continue
|
||||
}
|
||||
delta := capturedAt.Sub(observation.CapturedAt)
|
||||
delta := capturedAt.Sub(row.CapturedAt)
|
||||
if delta < 0 {
|
||||
delta = -delta
|
||||
}
|
||||
if delta > metricSnapshotOpenrestyMatchWindow {
|
||||
if delta > metricSnapshotEdgeHealthMatchWindow {
|
||||
continue
|
||||
}
|
||||
if matched == nil || delta < bestDelta {
|
||||
matched = observation
|
||||
matched = row
|
||||
bestDelta = delta
|
||||
}
|
||||
}
|
||||
@@ -676,21 +733,6 @@ func LatestMetricSnapshotsByNode(snapshots []*model.OpenFlareMetricSnapshot) map
|
||||
return result
|
||||
}
|
||||
|
||||
// LatestTrafficReportsByNode returns the latest traffic report per node.
|
||||
func LatestTrafficReportsByNode(reports []*model.OpenFlareRequestReport) map[string]*model.OpenFlareRequestReport {
|
||||
result := make(map[string]*model.OpenFlareRequestReport, len(reports))
|
||||
for _, report := range reports {
|
||||
if report == nil || report.NodeID == "" {
|
||||
continue
|
||||
}
|
||||
if existing, ok := result[report.NodeID]; ok && !report.WindowEndedAt.After(existing.WindowEndedAt) {
|
||||
continue
|
||||
}
|
||||
result[report.NodeID] = report
|
||||
}
|
||||
return result
|
||||
}
|
||||
|
||||
// ActiveHealthEventsByNode groups active health events by node id.
|
||||
func ActiveHealthEventsByNode(events []*model.OpenFlareHealthEvent) map[string][]*model.OpenFlareHealthEvent {
|
||||
result := make(map[string][]*model.OpenFlareHealthEvent)
|
||||
@@ -711,30 +753,6 @@ func Percentage(used int64, total int64) float64 {
|
||||
return (float64(used) / float64(total)) * percentageMultiplier
|
||||
}
|
||||
|
||||
func mergeJSONCounts(target distributionAccumulator, raw string) {
|
||||
if len(target) == 0 && strings.TrimSpace(raw) == "" {
|
||||
return
|
||||
}
|
||||
values := parseJSONCounts(raw)
|
||||
for key, value := range values {
|
||||
if strings.TrimSpace(key) == "" || value <= 0 {
|
||||
continue
|
||||
}
|
||||
target[key] += value
|
||||
}
|
||||
}
|
||||
|
||||
func parseJSONCounts(raw string) map[string]int64 {
|
||||
if strings.TrimSpace(raw) == "" {
|
||||
return nil
|
||||
}
|
||||
values := make(map[string]int64)
|
||||
if err := json.Unmarshal([]byte(raw), &values); err != nil {
|
||||
return nil
|
||||
}
|
||||
return values
|
||||
}
|
||||
|
||||
func toDistributionItems(values distributionAccumulator, limit int) []DistributionItem {
|
||||
if len(values) == 0 {
|
||||
return []DistributionItem{}
|
||||
|
||||
@@ -25,52 +25,25 @@ func TestBuildTrafficTrendPointsFromHourlyBucketsByHour(t *testing.T) {
|
||||
if len(points) != observabilityTrendBuckets {
|
||||
t.Fatalf("BuildTrafficTrendPointsFromHourly() len = %d, want %d", len(points), observabilityTrendBuckets)
|
||||
}
|
||||
}
|
||||
|
||||
func TestBuildTrafficTrendPointsBucketsByHour(t *testing.T) {
|
||||
t.Parallel()
|
||||
|
||||
now := time.Date(2026, 6, 19, 17, 30, 0, 0, time.UTC)
|
||||
reports := []*model.OpenFlareRequestReport{
|
||||
{
|
||||
NodeID: "node-a",
|
||||
WindowStartedAt: now.Add(-3 * time.Hour),
|
||||
WindowEndedAt: now.Add(-3*time.Hour + time.Minute),
|
||||
RequestCount: 10,
|
||||
ErrorCount: 1,
|
||||
},
|
||||
{
|
||||
NodeID: "node-a",
|
||||
WindowStartedAt: now.Add(-30 * time.Minute),
|
||||
WindowEndedAt: now.Add(-29 * time.Minute),
|
||||
RequestCount: 6,
|
||||
ErrorCount: 0,
|
||||
},
|
||||
}
|
||||
|
||||
points := BuildTrafficTrendPoints(now, reports)
|
||||
if len(points) != observabilityTrendBuckets {
|
||||
t.Fatalf("BuildTrafficTrendPoints() len = %d, want %d", len(points), observabilityTrendBuckets)
|
||||
}
|
||||
|
||||
var totalRequests int64
|
||||
// Hourly UV must not be summed into trend points.
|
||||
for _, point := range points {
|
||||
totalRequests += point.RequestCount
|
||||
if point.UniqueVisitorCount != 0 {
|
||||
t.Fatalf("UniqueVisitorCount = %d, want 0 on hourly path", point.UniqueVisitorCount)
|
||||
}
|
||||
}
|
||||
if totalRequests != 16 {
|
||||
t.Fatalf("total request_count = %d, want 16", totalRequests)
|
||||
index, ok := trendBucketIndex(now.Add(-2*time.Hour).Truncate(time.Hour), trendWindowStart(now))
|
||||
if !ok {
|
||||
t.Fatal("expected valid bucket index")
|
||||
}
|
||||
|
||||
currentHour := points[len(points)-1]
|
||||
if currentHour.RequestCount != 6 {
|
||||
t.Fatalf("current hour request_count = %d, want 6", currentHour.RequestCount)
|
||||
if points[index].RequestCount != 12 {
|
||||
t.Fatalf("request_count = %d, want 12", points[index].RequestCount)
|
||||
}
|
||||
if currentHour.ErrorCount != 0 {
|
||||
t.Fatalf("current hour error_count = %d, want 0", currentHour.ErrorCount)
|
||||
if points[index].ErrorCount != 1 {
|
||||
t.Fatalf("error_count = %d, want 1", points[index].ErrorCount)
|
||||
}
|
||||
}
|
||||
|
||||
func TestBuildMetricSnapshotViewsMergesOpenrestyObservation(t *testing.T) {
|
||||
func TestBuildMetricSnapshotViewsMergesEdgeHealthConnections(t *testing.T) {
|
||||
t.Parallel()
|
||||
|
||||
capturedAt := time.Date(2026, 6, 19, 12, 0, 0, 0, time.UTC)
|
||||
@@ -82,54 +55,30 @@ func TestBuildMetricSnapshotViewsMergesOpenrestyObservation(t *testing.T) {
|
||||
CPUUsagePercent: 12.5,
|
||||
},
|
||||
}
|
||||
openrestyObs := []*model.OpenFlareNodeObservationOpenresty{
|
||||
edgeHealth := []*model.OpenFlareEdgeHealth{
|
||||
{
|
||||
NodeID: "node-a",
|
||||
CapturedAt: capturedAt.Add(5 * time.Second),
|
||||
OpenrestyRxBytes: 4096,
|
||||
OpenrestyTxBytes: 8192,
|
||||
OpenrestyConnections: 7,
|
||||
NodeID: "node-a",
|
||||
CapturedAt: capturedAt.Add(5 * time.Second),
|
||||
Status: "healthy",
|
||||
Connections: 7,
|
||||
},
|
||||
}
|
||||
|
||||
views := BuildMetricSnapshotViews(snapshots, openrestyObs)
|
||||
views := BuildMetricSnapshotViews(snapshots, edgeHealth)
|
||||
if len(views) != 1 {
|
||||
t.Fatalf("BuildMetricSnapshotViews() len = %d, want 1", len(views))
|
||||
}
|
||||
if views[0].OpenrestyRxBytes != 4096 {
|
||||
t.Fatalf("OpenrestyRxBytes = %d, want 4096", views[0].OpenrestyRxBytes)
|
||||
}
|
||||
if views[0].OpenrestyTxBytes != 8192 {
|
||||
t.Fatalf("OpenrestyTxBytes = %d, want 8192", views[0].OpenrestyTxBytes)
|
||||
}
|
||||
if views[0].OpenrestyConnections != 7 {
|
||||
t.Fatalf("OpenrestyConnections = %d, want 7", views[0].OpenrestyConnections)
|
||||
}
|
||||
}
|
||||
|
||||
func TestLatestTrafficReportUsesLatestWindowEndedAt(t *testing.T) {
|
||||
func TestBuildTrafficWindowSummaryFromAccessLogsNilWithoutData(t *testing.T) {
|
||||
t.Parallel()
|
||||
|
||||
older := &model.OpenFlareRequestReport{
|
||||
WindowEndedAt: time.Date(2026, 6, 19, 10, 0, 0, 0, time.UTC),
|
||||
RequestCount: 3,
|
||||
}
|
||||
newer := &model.OpenFlareRequestReport{
|
||||
WindowEndedAt: time.Date(2026, 6, 19, 11, 0, 0, 0, time.UTC),
|
||||
RequestCount: 9,
|
||||
}
|
||||
|
||||
latest := latestTrafficReport([]*model.OpenFlareRequestReport{older, newer})
|
||||
if latest == nil || latest.RequestCount != 9 {
|
||||
t.Fatalf("latestTrafficReport() = %#v, want newer report with request_count 9", latest)
|
||||
}
|
||||
}
|
||||
|
||||
func TestBuildTrafficWindowSummaryNilWithoutReport(t *testing.T) {
|
||||
t.Parallel()
|
||||
|
||||
if summary := buildTrafficWindowSummary(nil); summary != nil {
|
||||
t.Fatalf("buildTrafficWindowSummary(nil) = %#v, want nil", summary)
|
||||
// Without an access-log store / data, summary is nil.
|
||||
if summary := buildTrafficWindowSummaryFromAccessLogs(t.Context(), "missing", time.Now().Add(-time.Hour), time.Now()); summary != nil {
|
||||
t.Fatalf("buildTrafficWindowSummaryFromAccessLogs() = %#v, want nil", summary)
|
||||
}
|
||||
}
|
||||
|
||||
@@ -173,12 +122,8 @@ func TestBuildNetworkTrendPointsUsesCounterDeltas(t *testing.T) {
|
||||
{NodeID: "n1", CapturedAt: base.Add(10 * time.Minute), NetworkRxBytes: 1000, NetworkTxBytes: 2000},
|
||||
{NodeID: "n1", CapturedAt: base.Add(20 * time.Minute), NetworkRxBytes: 1500, NetworkTxBytes: 2600},
|
||||
}
|
||||
openrestyObs := []*model.OpenFlareNodeObservationOpenresty{
|
||||
{NodeID: "n1", CapturedAt: base.Add(10 * time.Minute), OpenrestyRxBytes: 100, OpenrestyTxBytes: 200},
|
||||
{NodeID: "n1", CapturedAt: base.Add(20 * time.Minute), OpenrestyRxBytes: 180, OpenrestyTxBytes: 250},
|
||||
}
|
||||
|
||||
points := BuildNetworkTrendPoints(now, snapshots, openrestyObs)
|
||||
points := BuildNetworkTrendPoints(now, snapshots)
|
||||
current := points[len(points)-1]
|
||||
if current.NetworkRxBytes != 500 {
|
||||
t.Fatalf("network_rx_bytes = %d, want 500", current.NetworkRxBytes)
|
||||
@@ -186,11 +131,30 @@ func TestBuildNetworkTrendPointsUsesCounterDeltas(t *testing.T) {
|
||||
if current.NetworkTxBytes != 600 {
|
||||
t.Fatalf("network_tx_bytes = %d, want 600", current.NetworkTxBytes)
|
||||
}
|
||||
if current.OpenrestyRxBytes != 80 {
|
||||
t.Fatalf("openresty_rx_bytes = %d, want 80", current.OpenrestyRxBytes)
|
||||
if current.BytesReceived != 0 || current.BytesProvided != 0 {
|
||||
t.Fatalf("business bytes should be 0 without access logs overlay, got received=%d provided=%d",
|
||||
current.BytesReceived, current.BytesProvided)
|
||||
}
|
||||
if current.OpenrestyTxBytes != 50 {
|
||||
t.Fatalf("openresty_tx_bytes = %d, want 50", current.OpenrestyTxBytes)
|
||||
}
|
||||
|
||||
func TestBuildHealthSummaryUsesTrafficSummary(t *testing.T) {
|
||||
t.Parallel()
|
||||
|
||||
snapshot := &model.OpenFlareMetricSnapshot{
|
||||
CPUUsagePercent: 10,
|
||||
MemoryUsedBytes: 1,
|
||||
MemoryTotalBytes: 10,
|
||||
}
|
||||
traffic := &TrafficWindowSummary{
|
||||
RequestCount: 200,
|
||||
ErrorCount: 20, // 10% error rate
|
||||
}
|
||||
summary := buildHealthSummary(snapshot, traffic, nil)
|
||||
if !summary.HasTrafficRisk {
|
||||
t.Fatal("HasTrafficRisk = false, want true for 10% error rate with >=100 requests")
|
||||
}
|
||||
if summary.HasCapacityRisk {
|
||||
t.Fatal("HasCapacityRisk = true, want false")
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -79,7 +79,6 @@ type NodeView struct {
|
||||
NodeID string `json:"node_id"`
|
||||
Profile *model.OpenFlareNodeSystemProfile `json:"profile"`
|
||||
MetricSnapshots []*NodeMetricSnapshotView `json:"metric_snapshots"`
|
||||
TrafficReports []*model.OpenFlareRequestReport `json:"traffic_reports"`
|
||||
HealthEvents []*model.OpenFlareHealthEvent `json:"health_events"`
|
||||
Analytics NodeAnalytics `json:"analytics"`
|
||||
Trends NodeTrends `json:"trends"`
|
||||
@@ -118,11 +117,7 @@ func GetNodeObservability(ctx context.Context, id uint, query NodeQuery) (*NodeV
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
openrestyObs, err := model.ListOpenFlareNodeObservationOpenresty(ctx, node.NodeID, since, limit)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
reports, err := model.ListOpenFlareRequestReportsSince(ctx, node.NodeID, since, limit)
|
||||
edgeHealth, err := model.ListOpenFlareEdgeHealth(ctx, node.NodeID, since, limit)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
@@ -134,18 +129,21 @@ func GetNodeObservability(ctx context.Context, id uint, query NodeQuery) (*NodeV
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
distributions := BuildTrafficDistributionsFromAccessLogs(
|
||||
ctx, since, now, defaultTrafficDistributionLimit, accessLogRegions,
|
||||
)
|
||||
trafficSummary := buildTrafficWindowSummaryFromAccessLogs(ctx, node.NodeID, since, now)
|
||||
view := &NodeView{
|
||||
NodeID: node.NodeID,
|
||||
Profile: profile,
|
||||
MetricSnapshots: BuildMetricSnapshotViews(snapshots, openrestyObs),
|
||||
TrafficReports: reports,
|
||||
MetricSnapshots: BuildMetricSnapshotViews(snapshots, edgeHealth),
|
||||
HealthEvents: events,
|
||||
Analytics: NodeAnalytics{
|
||||
Traffic: buildTrafficWindowSummary(latestTrafficReport(reports)),
|
||||
Distributions: BuildTrafficDistributions(reports, accessLogRegions, defaultTrafficDistributionLimit),
|
||||
Health: buildHealthSummary(latestMetricSnapshot(snapshots), latestTrafficReport(reports), events),
|
||||
Traffic: trafficSummary,
|
||||
Distributions: distributions,
|
||||
Health: buildHealthSummary(latestMetricSnapshot(snapshots), trafficSummary, events),
|
||||
},
|
||||
Trends: BuildNodeTrends(ctx, now, node.NodeID, snapshots, openrestyObs, reports),
|
||||
Trends: BuildNodeTrends(ctx, now, node.NodeID, snapshots),
|
||||
}
|
||||
if node.NodeType == "tunnel_relay" {
|
||||
frpsObs, frpsErr := model.ListOpenFlareNodeObservationFrps(ctx, node.NodeID, time.Time{}, 1)
|
||||
|
||||
Reference in New Issue
Block a user