feat(aiops): 长窗口(>1h)分析必须指定应用
全命名空间 6h/24h 聚合数据量过大会拖垮 Loki:前端时间范围旁新增应用范围选择器 (labelValues 候选),>1h 未选应用禁用开始分析并提示;后端同样强制 400(闸门在服务端); 信号采集选择器按 app 收窄后 histogram/stats 从全量变为单应用流,秒级返回。
This commit is contained in:
opsx committed
1 parent
1610988c2e
commit
c090ba0564
7 files changed
+66
-14
No files matched your search
@@ -22,7 +22,7 @@ func (ctrl *AIController) GetConfig(c *gin.Context) {
|
||||
userID := c.GetUint("userId")
|
||||
cfg := services.MaskedAIConfig()
|
||||
utils.Success(c, gin.H{
|
||||
"config": cfg,
|
||||
"config": cfg,
|
||||
"tokenUsedToday": services.AITokenUsedToday(userID),
|
||||
})
|
||||
}
|
||||
@@ -91,6 +91,7 @@ func (ctrl *AIController) Analyze(c *gin.Context) {
|
||||
Start int64 `json:"start" binding:"required"`
|
||||
End int64 `json:"end" binding:"required"`
|
||||
Question string `json:"question"`
|
||||
App string `json:"app"`
|
||||
}
|
||||
if !utils.BindAndValidate(c, &req) {
|
||||
return
|
||||
@@ -110,13 +111,18 @@ func (ctrl *AIController) Analyze(c *gin.Context) {
|
||||
utils.BadRequest(c, fmt.Sprintf("时间范围过大（%d 分钟）：AI 分析最多支持 24 小时，建议用 15~60 分钟", int(to.Sub(from).Minutes())))
|
||||
return
|
||||
}
|
||||
// 长窗口必须指定应用:全命名空间聚合在 6h/24h 下数据量过大,查询会拖垮 Loki
|
||||
if to.Sub(from) > time.Hour && req.App == "" {
|
||||
utils.BadRequest(c, "超过 1 小时的分析必须指定应用(app):全命名空间长窗口聚合过慢,请先在应用排行里选一个 app 缩小范围")
|
||||
return
|
||||
}
|
||||
|
||||
userID := c.GetUint("userId")
|
||||
username := c.GetString("username")
|
||||
admin := c.GetString("role") == "admin"
|
||||
|
||||
cfg := services.GetAIConfig()
|
||||
outcome, err := services.RunAIAnalysis(c.Request.Context(), cfg, src, userID, username, admin, from, to, req.Question)
|
||||
outcome, err := services.RunAIAnalysis(c.Request.Context(), cfg, src, userID, username, admin, from, to, req.Question, req.App)
|
||||
if err != nil {
|
||||
// 未配置模型 / 配额用尽 / 上游报错：都给出具体原因，不包装成 500
|
||||
utils.Error(c, 400, err.Error())
|
||||
|
||||
@@ -59,7 +59,7 @@ JSON 结构：
|
||||
}`
|
||||
|
||||
// RunAIAnalysis 执行一次分析：采集信号 → 脱敏 → 调模型 → 解析 → 落审计。
|
||||
func RunAIAnalysis(ctx context.Context, cfg models.AIConfig, src *models.LogSource, userID uint, username string, admin bool, from, to time.Time, question string) (*AIAnalysisOutcome, error) {
|
||||
func RunAIAnalysis(ctx context.Context, cfg models.AIConfig, src *models.LogSource, userID uint, username string, admin bool, from, to time.Time, question, app string) (*AIAnalysisOutcome, error) {
|
||||
if !cfg.Enabled {
|
||||
return nil, fmt.Errorf("AI 分析未启用：请先在「智能运维 → 模型配置」中开启并保存")
|
||||
}
|
||||
@@ -79,7 +79,7 @@ func RunAIAnalysis(ctx context.Context, cfg models.AIConfig, src *models.LogSour
|
||||
}
|
||||
|
||||
started := time.Now()
|
||||
signals, err := CollectAISignals(ctx, src, userID, admin, from, to)
|
||||
signals, err := CollectAISignals(ctx, src, userID, admin, from, to, app)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
@@ -92,15 +92,20 @@ type AIAlert struct {
|
||||
}
|
||||
|
||||
// CollectAISignals 采集一次分析所需的全部信号。
|
||||
func CollectAISignals(ctx context.Context, src *models.LogSource, userID uint, admin bool, from, to time.Time) (*AISignalBundle, error) {
|
||||
func CollectAISignals(ctx context.Context, src *models.LogSource, userID uint, admin bool, from, to time.Time, app string) (*AISignalBundle, error) {
|
||||
if src == nil {
|
||||
return nil, fmt.Errorf("数据源不存在")
|
||||
}
|
||||
if !to.After(from) {
|
||||
return nil, fmt.Errorf("时间范围无效")
|
||||
}
|
||||
// 复用查询授权：AI 能看到的数据范围必须与当前用户一致，不多不少
|
||||
selector, err := AuthorizedLogQuery(src, userID, admin, `{namespace=~".+"}`)
|
||||
// 复用查询授权：AI 能看到的数据范围必须与当前用户一致，不多不少。
|
||||
// 长窗口由上层强制指定 app(缩小聚合范围);15m/1h 允许全命名空间
|
||||
baseSelector := `{namespace=~".+"}`
|
||||
if app != "" {
|
||||
baseSelector = `{namespace=~".+", app="` + app + `"}`
|
||||
}
|
||||
selector, err := AuthorizedLogQuery(src, userID, admin, baseSelector)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
@@ -84,7 +84,7 @@ export const aiApi = {
|
||||
saveConfig: (data: Partial<AIConfig>) => http.put<{ config: AIConfig }>('/ai/config', data),
|
||||
testConfig: (data: { baseUrl?: string; model?: string; apiKey?: string }) =>
|
||||
http.post<{ reply: string; model: string }>('/ai/config/test', data, { timeout: 45000 }),
|
||||
analyze: (data: { sourceId: number; start: number; end: number; question?: string }) =>
|
||||
analyze: (data: { sourceId: number; start: number; end: number; question?: string; app?: string }) =>
|
||||
http.post<AIAnalyzeResponse>('/ai/analyze', data, { timeout: 180000 }),
|
||||
runs: (page: number, pageSize: number) =>
|
||||
http.get<{ list: AIRun[]; total: number; page: number; pageSize: number }>(
|
||||
|
||||
@@ -9,6 +9,10 @@ export default {
|
||||
},
|
||||
|
||||
analyze: {
|
||||
appScope: 'App scope',
|
||||
allApps: 'All apps',
|
||||
appRequired: 'Required (over 1h)',
|
||||
appScopeHint: 'Analysis over 1 hour requires a specific app to keep aggregation fast',
|
||||
source: 'Data source',
|
||||
range: 'Time range',
|
||||
question: 'Extra question (optional)',
|
||||
|
||||
@@ -9,6 +9,10 @@ export default {
|
||||
},
|
||||
|
||||
analyze: {
|
||||
appScope: '应用范围',
|
||||
allApps: '全部应用',
|
||||
appRequired: '必选(超过 1 小时)',
|
||||
appScopeHint: '超过 1 小时的分析必须指定应用,避免全命名空间聚合过慢',
|
||||
source: '数据源',
|
||||
range: '时间范围',
|
||||
question: '补充问题（可选）',
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
import { useCallback, useEffect, useMemo, useState } from 'react'
|
||||
import { useCallback, useEffect, useState } from 'react'
|
||||
import { useTranslation } from 'react-i18next'
|
||||
import { LLM_PROVIDERS, matchProvider } from './llmProviders'
|
||||
import { Loader2, Play, Save, Sparkles, PlugZap, History, ShieldCheck } from 'lucide-react'
|
||||
@@ -50,12 +50,28 @@ export default function AiopsPage() {
|
||||
const [sources, setSources] = useState<LogSource[]>([])
|
||||
const [sourceId, setSourceId] = useState(0)
|
||||
const [rangeKey, setRangeKey] = useState<(typeof RANGE_PRESETS)[number]['key']>('15m')
|
||||
const [apps, setApps] = useState<string[]>([])
|
||||
const [app, setApp] = useState('')
|
||||
const [question, setQuestion] = useState('')
|
||||
const [running, setRunning] = useState(false)
|
||||
const [outcome, setOutcome] = useState<AIAnalyzeResponse | null>(null)
|
||||
const [promptOpen, setPromptOpen] = useState(false)
|
||||
const [enabled, setEnabled] = useState(false)
|
||||
|
||||
// app 候选:长窗口(>1h)必须指定应用收窄聚合范围
|
||||
useEffect(() => {
|
||||
setApp('')
|
||||
setApps([])
|
||||
if (!sourceId) return
|
||||
let active = true
|
||||
logsApi.labelValues(sourceId, 'app')
|
||||
.then(list => { if (active) setApps(Array.isArray(list) ? list : []) })
|
||||
.catch(() => { if (active) setApps([]) })
|
||||
return () => { active = false }
|
||||
}, [sourceId])
|
||||
const rangeMinutes = RANGE_PRESETS.find(r => r.key === rangeKey)?.minutes ?? 15
|
||||
const needApp = rangeMinutes > 60
|
||||
|
||||
// ── 配置 ──
|
||||
const [config, setConfig] = useState<AIConfig | null>(null)
|
||||
const [apiKeyInput, setApiKeyInput] = useState('')
|
||||
@@ -120,16 +136,16 @@ export default function AiopsPage() {
|
||||
if (tab === 'history') loadRuns(runPage)
|
||||
}, [tab, runPage, loadRuns])
|
||||
|
||||
const rangeMinutes = useMemo(
|
||||
() => RANGE_PRESETS.find(r => r.key === rangeKey)?.minutes ?? 15,
|
||||
[rangeKey],
|
||||
)
|
||||
|
||||
const runAnalysis = async () => {
|
||||
if (!sourceId) {
|
||||
toast.error('请先选择数据源')
|
||||
return
|
||||
}
|
||||
if (needApp && !app) {
|
||||
toast.error('超过 1 小时的分析必须指定应用：全命名空间长窗口聚合过慢')
|
||||
return
|
||||
}
|
||||
const end = Date.now()
|
||||
const start = end - rangeMinutes * 60 * 1000
|
||||
setRunning(true)
|
||||
@@ -140,6 +156,7 @@ export default function AiopsPage() {
|
||||
start: start * 1e6,
|
||||
end: end * 1e6,
|
||||
question: question.trim() || undefined,
|
||||
app: app || undefined,
|
||||
})
|
||||
setOutcome(res)
|
||||
loadConfig()
|
||||
@@ -255,6 +272,22 @@ export default function AiopsPage() {
|
||||
))}
|
||||
</div>
|
||||
</div>
|
||||
<div className="space-y-1">
|
||||
<Label className="text-xs">
|
||||
{t('analyze.appScope')}
|
||||
{needApp && <span className="ml-1 text-destructive">*</span>}
|
||||
</Label>
|
||||
<select
|
||||
aria-label={t('analyze.appScope')}
|
||||
className="h-9 min-w-40 rounded-md border bg-background px-3 text-sm"
|
||||
value={app}
|
||||
onChange={e => setApp(e.target.value)}
|
||||
>
|
||||
<option value="">{needApp ? t('analyze.appRequired') : t('analyze.allApps')}</option>
|
||||
{apps.map(a => <option key={a} value={a}>{a}</option>)}
|
||||
</select>
|
||||
{needApp && <p className="text-[11px] text-warning">{t('analyze.appScopeHint')}</p>}
|
||||
</div>
|
||||
<div className="min-w-64 flex-1 space-y-1">
|
||||
<Label className="text-xs">{t('analyze.question')}</Label>
|
||||
<Input
|
||||
@@ -264,7 +297,7 @@ export default function AiopsPage() {
|
||||
className="h-9"
|
||||
/>
|
||||
</div>
|
||||
<Button onClick={runAnalysis} disabled={running || !enabled} className="h-9">
|
||||
<Button onClick={runAnalysis} disabled={running || !enabled || (needApp && !app)} className="h-9">
|
||||
{running ? <Loader2 className="mr-1.5 h-4 w-4 animate-spin" /> : <Play className="mr-1.5 h-4 w-4" />}
|
||||
{running ? t('analyze.running') : t('analyze.run')}
|
||||
</Button>
|
||||
|
||||
Reference in new issue
Block a user