Перейти к основному содержимому

Operational Dashboard Catalog

Назначение

Каталог задает обязательные operational dashboards для production-эксплуатации Pin. Dashboard считается готовым только если у него есть владелец, SLO или рабочий порог, алерты, ссылка на runbook и bounded labels без PII, секретов, access tokens, raw provider payload и unbounded identifiers.

Общие Правила

  • Каждый backend service имеет service dashboard с RED: request rate, error rate, duration.
  • Все графики фильтруются минимум по environment, service, tenantTier, brandId там, где brandId имеет bounded cardinality.
  • Для provider flows допускаются labels providerCode, operation, status, errorCode; raw external reference в labels запрещен.
  • Для финансовых и security-critical панелей mandatory alerts являются blocking requirement для release.
  • Runbook хранится рядом с service/system документацией и содержит impact, первые проверки, mitigation, replay/reconciliation и escalation owner.
  • Dashboard не является source of truth для денег, KYC, incentive decision check или user limits; он показывает состояние source of truth и documented projections.

Обязательные Dashboards

DashboardOwnerЦельОсновные сигналы
platform-api-redPlatformЗдоровье gateway и backend API.http_requests_total, http_server_duration, http_errors_total, p95/p99 latency.
accounting-ledger-operationsAccountingКонтроль ledger postings, holds и reconciliation.posting latency, failed postings, imbalance detector, active holds, reconciliation gap.
payments-deposits-withdrawalsPaymentsДепозиты, withdrawals, webhooks и provider degradation.callback failures, pending withdrawals, provider latency, duplicate callbacks, payout failures.
gamehost-provider-callbacksGameHostИгровые session/round callbacks и settlement.bet/win/rollback latency, duplicate round events, settlement errors, provider timeout rate.
identity-securityIdentityAuth, sessions, MFA, permission changes и tenant isolation.login failures, MFA failures, suspicious sessions, permission mutations, scope violations.
kyc-aml-reviewKYCKYC queue, AML decisions и manual review backlog.queue size, review age, provider failure rate, decision latency, blocked withdrawals.
incentive-economy-guardrailsIncentiveIncentive cost, wagering, abuse и guardrail breaches.redemption errors, incentive liability, wagering stuck count, abuse flags, budget burn.
partner-traffic-payoutsPartnerPartner traffic quality, commissions и payouts.click/signup conversion, qualified FTD, commission pending, payout failures, fraud flags.
events-and-outboxPlatformKafka, MassTransit outbox/inbox и DLQ.outbox pending, publish delay, consumer lag, duplicate inbox count, DLQ growth.
data-platform-analyticsAnalyticsClickHouse/Elasticsearch freshness и report availability.ingestion delay, failed exports, freshness SLA, query latency, storage errors.
release-and-flagsPlatform/ReleaseRollout safety, risky flags и release health.active risky flags, kill switch activations, cohort error delta, release error budget burn.

RED Метрики

ОбластьRateErrorsDuration
APIhttp_requests_total by route, method, status5xx, controlled business failures by errorCodep50/p95/p99 http_server_duration
MediatR<service>_<command_or_query>_requests_total<service>_<command_or_query>_errors_total<service>_<command_or_query>_duration_ms
Provider adapter<service>_<provider>_<operation>_requests_totalsignature, timeout, validation, provider errorcallback/client p95/p99
Messagingconsumed/published messageshandler failures, DLQ, duplicatespublish delay, handler duration
Frontendpage/API call countJS errors, API problem responsespage load, interaction latency

Business Metrics

DomainМетрикиAlert Intent
Accountingledger transactions count, debit/credit sum, imbalance count, reconciliation gap, hold amount.Найти нарушение double-entry, зависшие holds и расхождение с projections.
Paymentsdeposit success rate, withdrawal approval age, provider callback failure rate, payout pending amount.Защитить денежные операции и SLA provider flows.
GameHostactive sessions, settled rounds, rollback ratio, provider duplicate count, jackpot settlement failures.Выявить broken callbacks, replay storms и settlement drift.
Incentiveactive incentive liability, incentive cost by brand, wagering completion rate, expired incentive amount.Ограничить incentive abuse и runaway cost.
Identity/KYClogin failure rate, account lockouts, KYC review age, AML hold count.Отследить fraud/security spikes и compliance backlog.
PartnerFTD quality, chargeback/refund impact, commission approval age, payout failure count.Не выплатить ошибочную commission и увидеть плохой traffic source.
Analyticsevent ingestion delay, report freshness, export failure rate.Не принимать решения по устаревшим данным.

SLO И Error Budget

SLOTargetDashboardAlert
Public API availability99.9% monthlyplatform-api-redburn rate 2% за 1 час или 5% за 6 часов.
Accounting posting availability99.95% monthlyaccounting-ledger-operationsлюбая серия failed postings выше порога 5 минут.
Accounting posting p95< 300 msaccounting-ledger-operationsp95 выше 300 ms 10 минут.
Provider callback p95< 500 msgamehost-provider-callbacksp95 выше 500 ms 10 минут по provider.
Deposit processing p95< 2 spayments-deposits-withdrawalsp95 выше 2 s 10 минут или spike failures.
Outbox publish delay p95< 30 sevents-and-outboxp95 выше 30 s 10 минут.
Critical report freshness< 15 mindata-platform-analyticsfreshness выше SLA 15 минут.

Alerts, Runbooks И Owners

AlertSeverityOwnerRunbook Requirements
accounting_ledger_imbalance_detectedCriticalAccounting + FinanceStop affected rollout, isolate tenant/brand, run ledger reconciliation, block direct balance repair.
payment_provider_callback_failure_spikeHighPaymentsCheck signatures, provider status, replay inbox, switch provider route only through approved flag.
gamehost_settlement_error_spikeHighGameHostPause provider cohort, inspect idempotency, reconcile game rounds with accounting postings.
tenant_scope_violation_detectedCriticalSecurityDisable actor/session, preserve audit evidence, notify security owner, review recent deploy/flag.
kyc_review_backlog_sla_breachMediumKYC/OpsRebalance review queue, fail closed for withdrawals needing KYC, record compliance note.
incentive_liability_guardrail_breachHighIncentive + FinanceDisable risky incentive flag, freeze new activations, reconcile liability and brand budget.
outbox_publish_delay_breachHighPlatformInspect DB/Kafka, avoid manual event inserts, resume publisher, verify idempotent consumers.
release_cohort_error_deltaHighRelease OwnerStop rollout, compare flagged/unflagged cohorts, decide rollback, kill switch or forward-fix.

Dashboard Definition Checklist

  • Название dashboard и owner team указаны явно.
  • Описаны RED metrics, business metrics и bounded labels.
  • Для critical path есть SLO или operational threshold.
  • Alert имеет severity, owner, runbook и escalation path.
  • Есть links на logs/traces по correlationId, tenantId, brandId, operation.
  • Нет PII, secrets, raw payment/KYC/provider data в labels, annotations и screenshots.
  • Для release dashboards есть фильтр по releaseVersion, gitSha, featureFlagKey, rolloutCohort.

Обратные ссылки

Автоссылки: где используется этот документ.