chore: establish local platform baseline

2026-08-26 06:41:54 +08:00
commit 29bdf4900f
575 changed files with 95644 additions and 0 deletions
@@ -0,0 +1,74 @@
+package orchestration
+
+import (
+	"context"
+	"errors"
+	"strings"
+	"time"
+
+	"github.com/jackc/pgx/v5/pgxpool"
+)
+
+var ErrProviderUnavailable = errors.New("provider is temporarily unavailable")
+
+type ProviderHealth struct {
+	pool             *pgxpool.Pool
+	failureThreshold int
+	cooldown         time.Duration
+	now              func() time.Time
+}
+
+func NewProviderHealth(pool *pgxpool.Pool, failureThreshold int, cooldown time.Duration) (*ProviderHealth, error) {
+	if pool == nil || failureThreshold < 1 || cooldown <= 0 {
+		return nil, errors.New("provider health requires a database, positive failure threshold and cooldown")
+	}
+	return &ProviderHealth{pool: pool, failureThreshold: failureThreshold, cooldown: cooldown, now: time.Now}, nil
+}
+
+func (h *ProviderHealth) AllowStart(ctx context.Context, tenantID, provider string) (bool, error) {
+	now := h.now().UTC()
+	var status string
+	var openUntil *time.Time
+	err := h.pool.QueryRow(ctx, `
+		INSERT INTO provider_dependency_state(tenant_id,provider,status,updated_at)
+		VALUES($1::uuid,$2,'AVAILABLE',$3)
+		ON CONFLICT (tenant_id,provider) DO UPDATE SET provider=EXCLUDED.provider
+		RETURNING status,open_until`, tenantID, provider, now).Scan(&status, &openUntil)
+	if err != nil {
+		return false, err
+	}
+	return status != "UNAVAILABLE" || openUntil == nil || !now.Before(*openUntil), nil
+}
+
+func (h *ProviderHealth) RecordFailure(ctx context.Context, tenantID, provider, errorCode string) error {
+	now := h.now().UTC()
+	code := strings.ToUpper(strings.TrimSpace(errorCode))
+	if code == "" {
+		code = "PROVIDER_TEMPORARY_ERROR"
+	}
+	_, err := h.pool.Exec(ctx, `
+		INSERT INTO provider_dependency_state(
+			tenant_id,provider,status,consecutive_failures,open_until,last_error_code,last_failure_at,updated_at
+		) VALUES(
+			$1::uuid,$2,CASE WHEN $3<=1 THEN 'UNAVAILABLE' ELSE 'AVAILABLE' END,1,
+			CASE WHEN $3<=1 THEN $4::timestamptz ELSE NULL END,$5,$6,$6
+		)
+		ON CONFLICT (tenant_id,provider) DO UPDATE SET
+			consecutive_failures=provider_dependency_state.consecutive_failures+1,
+			status=CASE WHEN provider_dependency_state.consecutive_failures+1 >= $3 THEN 'UNAVAILABLE' ELSE provider_dependency_state.status END,
+			open_until=CASE WHEN provider_dependency_state.consecutive_failures+1 >= $3 THEN $4::timestamptz ELSE provider_dependency_state.open_until END,
+			last_error_code=$5,last_failure_at=$6,row_version=provider_dependency_state.row_version+1,updated_at=$6`,
+		tenantID, provider, h.failureThreshold, now.Add(h.cooldown), code, now)
+	return err
+}
+
+func (h *ProviderHealth) RecordSuccess(ctx context.Context, tenantID, provider string) error {
+	now := h.now().UTC()
+	_, err := h.pool.Exec(ctx, `
+		INSERT INTO provider_dependency_state(tenant_id,provider,status,last_success_at,updated_at)
+		VALUES($1::uuid,$2,'AVAILABLE',$3,$3)
+		ON CONFLICT (tenant_id,provider) DO UPDATE SET
+			status='AVAILABLE',consecutive_failures=0,open_until=NULL,last_error_code=NULL,
+			last_success_at=$3,row_version=provider_dependency_state.row_version+1,updated_at=$3`, tenantID, provider, now)
+	return err
+}