Precompute dashboard metrics and tune production workload

.env.example

@@ -54,6 +54,17 @@ QUERY_LIST_CACHE_MAX_ENTRIES=4
 GLOBAL_TREND_CACHE_MAX_ENTRIES=64
 QUERY_LIST_CACHE_MAX_ROWS=100000
 QUERY_LIST_CACHE_MAX_BYTES=67108864
+# Kalici query-metrics materialized snapshotlarini ayri worker yeniler. API
+# request yolunda advisor.query_metrics() calismaz. Concurrent refresh boyunca
+# onceki tamamlanmis snapshot servis edilmeye devam eder.
+QUERY_METRICS_SNAPSHOT_POLL_SECONDS=15
+QUERY_METRICS_SNAPSHOT_1H_REFRESH_SECONDS=900
+QUERY_METRICS_SNAPSHOT_24H_REFRESH_SECONDS=3600
+QUERY_METRICS_SNAPSHOT_7D_REFRESH_SECONDS=21600
+QUERY_METRICS_SNAPSHOT_30D_REFRESH_SECONDS=43200
+QUERY_METRICS_SNAPSHOT_STATEMENT_TIMEOUT_SECONDS=1800
+QUERY_METRICS_SNAPSHOT_RETRY_SECONDS=60
+QUERY_METRICS_SNAPSHOT_WORKER_MEMORY_LIMIT=256m
 # Tam snapshotlarin Python nesne ek yükü ayrıca container seviyesinde sınırlıdır.
 API_MEMORY_LIMIT=1g
 # Eszamanli repository migration runner'larinin advisory lock bekleme siniri.
@@ -91,6 +102,10 @@ WORKLOAD_PROFILE=normal
 WORKLOAD_DURATION_SECONDS=0
 WORKLOAD_WORKERS=6
 WORKLOAD_INTERVAL_SECONDS=0.25
+WORKLOAD_INTERVAL_JITTER_RATIO=0
+WORKLOAD_TRAFFIC_PHASE_SECONDS=0
+WORKLOAD_TRAFFIC_MIN_INTERVAL_MULTIPLIER=1
+WORKLOAD_TRAFFIC_MAX_INTERVAL_MULTIPLIER=1
 WORKLOAD_REPORT_INTERVAL_SECONDS=10
 WORKLOAD_RANDOM_SEED=20260725
 WORKLOAD_STATEMENT_TIMEOUT_MS=15000

.github/workflows/postgres-integration.yml

@@ -95,15 +95,18 @@ jobs:
 
       - name: Verify health and migration idempotency
         run: |
-          curl --fail --silent --show-error \
-            http://127.0.0.1:8000/api/v1/health \
-            | python -c '
-          import json, sys
-          health = json.load(sys.stdin)
-          assert health["status"] == "healthy", health
-          assert health["repository"] == "healthy", health
-          assert health["collector"] == "healthy", health
-          '
+          health_ready=false
+          for attempt in $(seq 1 18); do
+            if curl --fail --silent --show-error \
+              http://127.0.0.1:8000/api/v1/health \
+              | python -c 'import json, sys; health = json.load(sys.stdin); assert health["status"] == "healthy", health; assert health["repository"] == "healthy", health; assert health["collector"] == "healthy", health'; then
+              health_ready=true
+              break
+            fi
+            printf 'collector health attempt=%s is not ready yet\n' "$attempt"
+            sleep 10
+          done
+          test "$health_ready" = true
           docker compose run --rm repository-migrate
           bash scripts/verify-temporal-reliability.sh
           docker compose exec -T repository-db \
@@ -118,7 +121,7 @@ jobs:
             psql -X --set=ON_ERROR_STOP=1 --username postgres --dbname powa --file=- \
             < sql/tests/join_outbox_guardrail_integration.sql
 
-      - name: Verify a persisted 0013 database upgrades to 0014
+      - name: Verify a persisted 0013 database upgrades through 0016
         shell: bash
         run: |
           set -Eeuo pipefail
@@ -170,7 +173,7 @@ jobs:
               release record;
           BEGIN
               SELECT * INTO STRICT release FROM advisor.release_info();
-              IF release.current_migration <> '0014' OR release.applied_count <> 14 THEN
+              IF release.current_migration <> '0016' OR release.applied_count <> 16 THEN
                   RAISE EXCEPTION 'unexpected upgraded release state: %', row_to_json(release);
               END IF;
               IF NOT has_function_privilege(
@@ -207,6 +210,27 @@ jobs:
       - name: Prepare the deterministic runtime candidate
         run: bash scripts/verify.sh
 
+      - name: Refresh the disposable 30d runtime replay snapshot
+        run: |
+          docker compose exec -T query-metrics-snapshot-worker python - <<'PY'
+          import time
+
+          from app.config import get_settings
+          from app.snapshot_worker import open_connection, refresh_snapshot
+
+
+          settings = get_settings()
+          with open_connection(settings) as connection:
+              for _ in range(30):
+                  if refresh_snapshot(connection, "30d"):
+                      break
+                  time.sleep(2)
+              else:
+                  raise SystemExit(
+                      "30d query metrics snapshot advisory lock could not be acquired"
+                  )
+          PY
+
       - name: Boot disposable clone services
         run: |
           docker compose --profile real-validation up -d --wait \

CHANGELOG.md

@@ -4,6 +4,15 @@ Bu proje [Semantic Versioning](https://semver.org/) kullanır. Repository
 migration'ları ileri yönlüdür; uygulama sürümü ile şema uyumluluğu yükseltme
 runbook'unda belirtilir.
 
+## 1.1.1 — 2026-07-30
+
+- 1h/24h/7d/30d query-metrics sonuçları kalıcı materialized snapshot olarak
+  request yolundan önce hesaplanır.
+- Ayrı worker snapshot'ları sırayla ve atomik olarak yeniler; hesaplama sürerken
+  önceki tamamlanmış sonuç servis edilir.
+- Repository migration hedefi `0016`; query-metrics snapshot'larına ek olarak
+  global/server/database overview trendleri de önceden hesaplanır.
+
 ## 1.1.0 — 2026-07-26
 
 - Ana kaynakta yalnız persisted, salt-okunur sorgular için gerçek `EXPLAIN ANALYZE`.

README.md

@@ -1,6 +1,6 @@
 # PostgreSQL Sorgu Performansı ve Öneri Motoru
 
-Güncel sürüm: `1.1.0`. Değişiklikler [CHANGELOG.md](CHANGELOG.md), güvenli
+Güncel sürüm: `1.1.1`. Değişiklikler [CHANGELOG.md](CHANGELOG.md), güvenli
 yükseltme ve geri dönüş adımları [upgrade/rollback runbook'unda](docs/UPGRADE_ROLLBACK.md).
 
 PDF v1.1'de tarif edilen ilk iterasyonun çalışan referans uygulamasıdır. Tek bir Docker/OrbStack hostu üzerinde **iki ayrı PostgreSQL sunucu süreci** çalışır: demo kaynak instance `5432`, PoWA repository instance `5433`. PoWA Collector istatistikleri kaynaktan repository'ye taşır; FastAPI yalnız repository'yi okur ve React arayüzü sonuçları gösterir. Aynı repository/collector, `scripts/register-source.sh` ile birden fazla gerçek PostgreSQL kaynağı izleyebilir.

VERSION

@@ -1 +1 @@
-1.1.0
+1.1.1

backend/app/config.py

@@ -24,6 +24,20 @@
     "30d": "1 day",
 }
 
+QUERY_METRICS_SNAPSHOT_VIEWS: dict[str, str] = {
+    "1h": "query_metrics_snapshot_1h",
+    "24h": "query_metrics_snapshot_24h",
+    "7d": "query_metrics_snapshot_7d",
+    "30d": "query_metrics_snapshot_30d",
+}
+
+GLOBAL_TREND_SNAPSHOT_VIEWS: dict[str, str] = {
+    "1h": "global_trend_snapshot_1h",
+    "24h": "global_trend_snapshot_24h",
+    "7d": "global_trend_snapshot_7d",
+    "30d": "global_trend_snapshot_30d",
+}
+
 
 PrincipalRole = Literal["analyst", "annotator", "admin"]
 
@@ -91,6 +105,26 @@ class Settings(BaseSettings):
         ge=1024 * 1024,
         le=1024 * 1024 * 1024,
     )
+    # A dedicated worker refreshes persistent materialized snapshots.  API
+    # requests only read those snapshots and therefore never execute the
+    # expensive query_metrics function on a cold process-local cache.
+    query_metrics_snapshot_poll_seconds: float = Field(default=15.0, ge=1, le=300)
+    query_metrics_snapshot_1h_refresh_seconds: int = Field(
+        default=15 * 60, ge=60, le=86_400
+    )
+    query_metrics_snapshot_24h_refresh_seconds: int = Field(
+        default=60 * 60, ge=60, le=7 * 86_400
+    )
+    query_metrics_snapshot_7d_refresh_seconds: int = Field(
+        default=6 * 60 * 60, ge=60, le=30 * 86_400
+    )
+    query_metrics_snapshot_30d_refresh_seconds: int = Field(
+        default=12 * 60 * 60, ge=60, le=30 * 86_400
+    )
+    query_metrics_snapshot_statement_timeout_seconds: int = Field(
+        default=30 * 60, ge=60, le=6 * 60 * 60
+    )
+    query_metrics_snapshot_retry_seconds: int = Field(default=60, ge=5, le=3_600)
     sql_text_visibility: str = "authorized"
     retention_days: int = 90
     log_level: str = "INFO"

backend/app/main.py

@@ -7,12 +7,15 @@
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
 from prometheus_client import CONTENT_TYPE_LATEST, generate_latest
+from starlette import status
+from starlette.requests import Request
+from starlette.responses import JSONResponse
 from starlette.responses import Response
 
 from app.api.router import router
 from app.config import get_settings
 from app.db import close_pool, open_pool
-from app.repositories.powa import repository
+from app.repositories.powa import QueryMetricsSnapshotWarming, repository
 from app.version import APPLICATION_VERSION
 
 
@@ -58,6 +61,18 @@ async def lifespan(_: FastAPI):
 app.include_router(router)
 
 
+@app.exception_handler(QueryMetricsSnapshotWarming)
+async def snapshot_warming(
+    _: Request,
+    __: QueryMetricsSnapshotWarming,
+) -> JSONResponse:
+    return JSONResponse(
+        status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
+        content={"detail": "Dashboard verileri ilk kez hazirlaniyor; kisa sure sonra deneyin."},
+        headers={"Retry-After": "30"},
+    )
+
+
 @app.get("/", include_in_schema=False)
 async def root() -> dict[str, str]:
     return {"name": app.title, "docs": "/docs", "health": "/api/v1/health"}