345a965fa3
功能特性： - 自动识别并标记 count_tokens 请求的错误 - 支持配置是否在统计中忽略 count_tokens 错误 - 错误数据完整保留，仅在统计时动态过滤 技术实现： - ops_error_logger.go: 自动标记 count_tokens 请求 - ops_repo.go: INSERT 语句添加 is_count_tokens 字段 - ops_repo_dashboard.go: buildErrorWhere 核心过滤函数 - ops_repo_preagg.go: 预聚合统计中添加过滤 - ops_repo_trends.go: 趋势统计查询添加过滤（2 处） - ops_settings_models.go: 添加 ignore_count_tokens_errors 配置 - ops_settings.go: 配置验证和默认值设置 - ops_port.go: 错误日志模型添加 IsCountTokens 字段 业务价值： - count_tokens 是探测性请求，其错误不影响真实业务 SLA - 用户可根据需求灵活控制是否计入统计 - 提升错误率、告警等运维指标的准确性 影响范围： - Dashboard 概览统计 - 错误趋势图表 - 告警规则评估 - 预聚合指标（hourly/daily） - 健康分数计算
100 lines
4.3 KiB
Go
100 lines
4.3 KiB
Go
package service
|
|
|
|
// Ops settings models stored in DB `settings` table (JSON blobs).
|
|
|
|
type OpsEmailNotificationConfig struct {
|
|
Alert OpsEmailAlertConfig `json:"alert"`
|
|
Report OpsEmailReportConfig `json:"report"`
|
|
}
|
|
|
|
type OpsEmailAlertConfig struct {
|
|
Enabled bool `json:"enabled"`
|
|
Recipients []string `json:"recipients"`
|
|
MinSeverity string `json:"min_severity"`
|
|
RateLimitPerHour int `json:"rate_limit_per_hour"`
|
|
BatchingWindowSeconds int `json:"batching_window_seconds"`
|
|
IncludeResolvedAlerts bool `json:"include_resolved_alerts"`
|
|
}
|
|
|
|
type OpsEmailReportConfig struct {
|
|
Enabled bool `json:"enabled"`
|
|
Recipients []string `json:"recipients"`
|
|
DailySummaryEnabled bool `json:"daily_summary_enabled"`
|
|
DailySummarySchedule string `json:"daily_summary_schedule"`
|
|
WeeklySummaryEnabled bool `json:"weekly_summary_enabled"`
|
|
WeeklySummarySchedule string `json:"weekly_summary_schedule"`
|
|
ErrorDigestEnabled bool `json:"error_digest_enabled"`
|
|
ErrorDigestSchedule string `json:"error_digest_schedule"`
|
|
ErrorDigestMinCount int `json:"error_digest_min_count"`
|
|
AccountHealthEnabled bool `json:"account_health_enabled"`
|
|
AccountHealthSchedule string `json:"account_health_schedule"`
|
|
AccountHealthErrorRateThreshold float64 `json:"account_health_error_rate_threshold"`
|
|
}
|
|
|
|
// OpsEmailNotificationConfigUpdateRequest allows partial updates, while the
|
|
// frontend can still send the full config shape.
|
|
type OpsEmailNotificationConfigUpdateRequest struct {
|
|
Alert *OpsEmailAlertConfig `json:"alert"`
|
|
Report *OpsEmailReportConfig `json:"report"`
|
|
}
|
|
|
|
type OpsDistributedLockSettings struct {
|
|
Enabled bool `json:"enabled"`
|
|
Key string `json:"key"`
|
|
TTLSeconds int `json:"ttl_seconds"`
|
|
}
|
|
|
|
type OpsAlertSilenceEntry struct {
|
|
RuleID *int64 `json:"rule_id,omitempty"`
|
|
Severities []string `json:"severities,omitempty"`
|
|
|
|
UntilRFC3339 string `json:"until_rfc3339"`
|
|
Reason string `json:"reason"`
|
|
}
|
|
|
|
type OpsAlertSilencingSettings struct {
|
|
Enabled bool `json:"enabled"`
|
|
|
|
GlobalUntilRFC3339 string `json:"global_until_rfc3339"`
|
|
GlobalReason string `json:"global_reason"`
|
|
|
|
Entries []OpsAlertSilenceEntry `json:"entries,omitempty"`
|
|
}
|
|
|
|
type OpsMetricThresholds struct {
|
|
SLAPercentMin *float64 `json:"sla_percent_min,omitempty"` // SLA低于此值变红
|
|
LatencyP99MsMax *float64 `json:"latency_p99_ms_max,omitempty"` // 延迟P99高于此值变红
|
|
TTFTp99MsMax *float64 `json:"ttft_p99_ms_max,omitempty"` // TTFT P99高于此值变红
|
|
RequestErrorRatePercentMax *float64 `json:"request_error_rate_percent_max,omitempty"` // 请求错误率高于此值变红
|
|
UpstreamErrorRatePercentMax *float64 `json:"upstream_error_rate_percent_max,omitempty"` // 上游错误率高于此值变红
|
|
}
|
|
|
|
type OpsAlertRuntimeSettings struct {
|
|
EvaluationIntervalSeconds int `json:"evaluation_interval_seconds"`
|
|
|
|
DistributedLock OpsDistributedLockSettings `json:"distributed_lock"`
|
|
Silencing OpsAlertSilencingSettings `json:"silencing"`
|
|
Thresholds OpsMetricThresholds `json:"thresholds"` // 指标阈值配置
|
|
}
|
|
|
|
// OpsAdvancedSettings stores advanced ops configuration (data retention, aggregation).
|
|
type OpsAdvancedSettings struct {
|
|
DataRetention OpsDataRetentionSettings `json:"data_retention"`
|
|
Aggregation OpsAggregationSettings `json:"aggregation"`
|
|
IgnoreCountTokensErrors bool `json:"ignore_count_tokens_errors"`
|
|
AutoRefreshEnabled bool `json:"auto_refresh_enabled"`
|
|
AutoRefreshIntervalSec int `json:"auto_refresh_interval_seconds"`
|
|
}
|
|
|
|
type OpsDataRetentionSettings struct {
|
|
CleanupEnabled bool `json:"cleanup_enabled"`
|
|
CleanupSchedule string `json:"cleanup_schedule"`
|
|
ErrorLogRetentionDays int `json:"error_log_retention_days"`
|
|
MinuteMetricsRetentionDays int `json:"minute_metrics_retention_days"`
|
|
HourlyMetricsRetentionDays int `json:"hourly_metrics_retention_days"`
|
|
}
|
|
|
|
type OpsAggregationSettings struct {
|
|
AggregationEnabled bool `json:"aggregation_enabled"`
|
|
}
|