opensource/cloudpods
3
0
Fork 0
mirror of https://github.com/yunionio/cloudpods.git synced 2026-09-24 16:03:43 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix(monitor): top query of monitor_resource_alert (#24052)

Browse Source
This commit is contained in:
Zexi Li
2026-01-09 12:10:17 +08:00
committed by GitHub
parent 97becd57ba
commit 7bd4869998
2 changed files with 68 additions and 48 deletions
Download Patch File Download Diff File Expand all files Collapse all files
pkg/mcclient/modules/monitor/mod_monitor_resource.go
+1 -1
View File
@@ -51,7 +51,7 @@ func NewMonitorResourceManager() *SMonitorResourceManager {
func newAlertResourceAlertManager() *SMonitorResourceAlertManager {
man := modules.NewJointMonitorV2Manager("monitorresourcealert", "monitorresourcealerts",
[]string{"monitor_resource_id", "alert_id", "res_name", "res_type", "metric", "alert_name", "alert_state", "send_state", "level",
[]string{"monitor_resource_id", "alert_count", "alert_id", "res_name", "res_type", "metric", "alert_name", "alert_state", "send_state", "level",
"trigger_time", "data"},
[]string{},
MonitorResourceManager, CommonAlerts)
pkg/monitor/models/monitor_resource_alert.go
+67 -47
View File
@@ -204,7 +204,6 @@ func (m *SMonitorResourceAlertManager) GetNowAlertingAlerts(ctx context.Context,
func (m *SMonitorResourceAlertManager) ListItemFilter(ctx context.Context, q *sqlchemy.SQuery, userCred mcclient.TokenCredential, input *monitor.MonitorResourceJointListInput) (*sqlchemy.SQuery, error) {
// 如果指定了时间段、top 和 alert_id 参数，执行特殊的 top 查询
// 使用 RawQuery 以包含 deleted 的数据（已恢复的资源）
if input.Top != nil {
// 加上 top 的参数校验
if input.Top == nil || *input.Top <= 0 {
@@ -216,9 +215,6 @@ func (m *SMonitorResourceAlertManager) ListItemFilter(ctx context.Context, q *sq
if input.StartTime.After(input.EndTime) {
return nil, httperrors.NewInputParameterError("start_time must be before end_time")
}
if len(input.AlertId) == 0 {
return nil, httperrors.NewInputParameterError("alert_id must be specified")
}
return m.getTopResourcesByMetricAndAlertCount(ctx, q, userCred, input)
}
@@ -302,7 +298,6 @@ func (m *SMonitorResourceAlertManager) CustomizeFilterList(ctx context.Context,
}
// getTopResourcesByMetricAndAlertCount 查询指定时间段内，某个监控策略下各监控指标报警资源最多的 top N 资源
// 使用 RawQuery 查询包含 deleted 的数据，以包含已恢复的资源
func (m *SMonitorResourceAlertManager) getTopResourcesByMetricAndAlertCount(
ctx context.Context,
q *sqlchemy.SQuery,
@@ -315,9 +310,11 @@ func (m *SMonitorResourceAlertManager) getTopResourcesByMetricAndAlertCount(
return nil, err
}
// 查询指定时间段内的 AlertRecord，过滤 alert_id
recordQuery := AlertRecordManager.Query("id", "alert_rule", "res_ids", "res_type")
recordQuery = recordQuery.Equals("alert_id", input.AlertId)
// 查询指定时间段内的 AlertRecord，获取 alert_id、res_ids 和 alert_rule（用于解析 metric）
recordQuery := AlertRecordManager.Query("id", "alert_id", "res_ids", "res_type", "alert_rule")
if len(input.AlertId) > 0 {
recordQuery = recordQuery.Equals("alert_id", input.AlertId)
}
recordQuery = recordQuery.GE("created_at", startTime).LE("created_at", endTime)
recordQuery = recordQuery.IsNotEmpty("res_ids")
@@ -329,9 +326,10 @@ func (m *SMonitorResourceAlertManager) getTopResourcesByMetricAndAlertCount(
// 执行查询获取所有记录
type RecordRow struct {
Id string
AlertRule jsonutils.JSONObject
AlertId string
ResIds string
ResType string
AlertRule jsonutils.JSONObject
}
rows := make([]RecordRow, 0)
err = recordQuery.All(&rows)
@@ -339,9 +337,15 @@ func (m *SMonitorResourceAlertManager) getTopResourcesByMetricAndAlertCount(
return nil, errors.Wrap(err, "query alert records")
}
// 按 metric 分组统计，然后合并所有 metric 的统计结果
// metricResourceCount[metric][resId] = count
metricResourceCount := make(map[string]map[string]int)
// 按照 resId、alert_id 和 metric 三个维度统计告警数量
// resourceAlertMetricCount[resId][alertId][metric] = count
type ResourceAlertMetricKey struct {
ResId string
AlertId string
Metric string
}
resourceAlertMetricCount := make(map[ResourceAlertMetricKey]int)
for _, row := range rows {
if len(row.ResIds) == 0 {
continue
@@ -369,71 +373,87 @@ func (m *SMonitorResourceAlertManager) getTopResourcesByMetricAndAlertCount(
if len(input.ResType) > 0 && row.ResType != input.ResType {
continue
}
// 对于每个 metric，统计资源数量
// 对于每个 metric，统计 (resId, alertId, metric) 组合的数量
for _, rule := range alertRules {
if len(rule.Metric) == 0 {
continue
}
if metricResourceCount[rule.Metric] == nil {
metricResourceCount[rule.Metric] = make(map[string]int)
key := ResourceAlertMetricKey{
ResId: resId,
AlertId: row.AlertId,
Metric: rule.Metric,
}
metricResourceCount[rule.Metric][resId]++
resourceAlertMetricCount[key]++
}
}
}
// 合并所有 metric 的统计结果，计算每个资源的总报警数
resourceCount := make(map[string]int)
for _, resourceCountByMetric := range metricResourceCount {
for resId, count := range resourceCountByMetric {
resourceCount[resId] += count
}
}
log.Infof("=======resourceCount: %#v", resourceCount)
// 转换为切片并按报警数量排序
type ResourceCount struct {
ResId string
Count int
type ResourceAlertMetricCount struct {
ResId string
AlertId string
Metric string
Count int
}
resourceCounts := make([]ResourceCount, 0, len(resourceCount))
for resId, count := range resourceCount {
resourceCounts = append(resourceCounts, ResourceCount{
ResId: resId,
Count: count,
counts := make([]ResourceAlertMetricCount, 0, len(resourceAlertMetricCount))
for key, count := range resourceAlertMetricCount {
counts = append(counts, ResourceAlertMetricCount{
ResId: key.ResId,
AlertId: key.AlertId,
Metric: key.Metric,
Count: count,
})
}
// 按报警数量降序排序
for i := 0; i < len(resourceCounts)-1; i++ {
for j := i + 1; j < len(resourceCounts); j++ {
if resourceCounts[i].Count < resourceCounts[j].Count {
resourceCounts[i], resourceCounts[j] = resourceCounts[j], resourceCounts[i]
for i := 0; i < len(counts)-1; i++ {
for j := i + 1; j < len(counts); j++ {
if counts[i].Count < counts[j].Count {
counts[i], counts[j] = counts[j], counts[i]
}
}
}
// 获取全局 top N 的资源 ID
topResIds := make([]string, 0, top)
for i := 0; i < min(top, len(resourceCounts)); i++ {
topResIds = append(topResIds, resourceCounts[i].ResId)
// 获取全局 top N 的 (resId, alertId, metric) 组合
topN := min(top, len(counts))
topCombinations := make([]ResourceAlertMetricCount, 0, topN)
for i := 0; i < topN; i++ {
topCombinations = append(topCombinations, counts[i])
}
log.Infof("top %d resources: %v", top, resourceCounts[:min(top, len(resourceCounts))])
log.Infof("top %d combinations: %v", top, topCombinations)
log.Infof("====topResIds: %#v", topResIds)
if len(topResIds) == 0 {
if len(topCombinations) == 0 {
// 如果没有找到任何记录，返回空查询
return q.FilterByFalse(), nil
}
q = m.RawQuery()
q = q.Equals("alert_id", input.AlertId)
q = q.Filter(sqlchemy.In(q.Field("monitor_resource_id"), topResIds))
// 构建查询条件：匹配 top N 的 (resId, alertId, metric) 组合
q = m.Query()
conditions := make([]sqlchemy.ICondition, 0, len(topCombinations))
for _, combo := range topCombinations {
cond := sqlchemy.AND(
sqlchemy.Equals(q.Field("monitor_resource_id"), combo.ResId),
sqlchemy.Equals(q.Field("alert_id"), combo.AlertId),
sqlchemy.Equals(q.Field("metric"), combo.Metric),
)
conditions = append(conditions, cond)
}
q = q.Filter(sqlchemy.OR(conditions...))
// 应用其他过滤条件
if len(input.AlertState) > 0 {
q = q.Equals("alert_state", input.AlertState)
}
if input.Alerting {
q = q.Equals("alert_state", monitor.AlertStateAlerting)
resQ := MonitorResourceManager.Query("res_id")
resQ, err = MonitorResourceManager.ListItemFilter(ctx, resQ, userCred, monitor.MonitorResourceListInput{})
if err != nil {
return q, errors.Wrap(err, "Get monitor in Query err")
}
resQ = m.SMonitorScopedResourceManager.FilterByOwner(ctx, resQ, m, userCred, userCred, rbacscope.TRbacScope(input.Scope))
q = q.Filter(sqlchemy.In(q.Field("monitor_resource_id"), resQ.SubQuery()))
}
if len(input.SendState) != 0 {
q = q.Equals("send_state", input.SendState)
}