EvoBGP/internal/httpapi/server.go

package httpapi

import (
	"context"
	"crypto/ed25519"
	"crypto/rand"
	"encoding/hex"
	"errors"
	"net/http"
	"strings"
	"sync"
	"time"

	"evobgp/internal/audit"
	"evobgp/internal/jobs"
	"evobgp/internal/logging"
	"evobgp/internal/maintenance"
	"evobgp/internal/pgmonitor"
	"evobgp/internal/runtimelogs"
	"evobgp/internal/store"
	"evobgp/internal/telemetry"

	"github.com/jackc/pgx/v5/pgxpool"
)

// Server implements EvoBGP control-plane HTTP API.
type Server struct {
	store                   store.Backend
	pgPool                  *pgxpool.Pool
	pgMonitor               *pgmonitor.Service
	maintConfig             *maintenance.ConfigProvider
	maintStats              *maintenance.DBStatsProvider
	jobs                    *jobs.Registry
	authLimiter             *authRateLimiter
	operationOnce           sync.Once
	operationLimiter        *operationLimiter
	bundlePriv              ed25519.PrivateKey
	bundleCache             *nodeBundleCache
	nodeBundleBuild         nodeBundleBuilder
	keyResolver             *apiKeyResolver
	firewallResolver        *firewallTokenResolver
	bundleSeedHex           string
	corsOrigins             []string
	cdnHTTP                 *http.Client
	runtimeLogs             *runtimelogs.Service
	runtimeLogsPolicyTenant string
	mux                     *http.ServeMux
	backgroundMu            sync.Mutex
	backgroundStarted       bool
	backgroundClosed        bool
	backgroundCancel        context.CancelFunc
	telemetryCollector      *telemetry.Collector

	// Portal / dual-auth (JWT) configuration.
	jwtSecret         string
	authIssuer        string
	authPortalURL     string
	portalTenantID    string
	authRequired      bool
	auditIngestSecret string
	auditPusher       *audit.PortalPusher
}

// Options configures the API server.
type Options struct {
	APIKeys string
	// DatabaseURL enables PostgreSQL-backed store (migrations applied on connect).
	DatabaseURL        string
	InsecureDev        bool
	SeedDemo           bool
	BundleSeedHex      string
	CORSAllowedOrigins string
	// RuntimeLogsPolicyTenant overrides tenant for auto-cleanup scheduler settings (optional).
	RuntimeLogsPolicyTenant string

	// Portal / dual-auth (JWT) — leave empty to disable JWT path.
	JWTSecret      string // AUTH_JWT_SECRET / EVOBGP_AUTH_JWT_SECRET (HS256 shared secret)
	AuthIssuer     string // AUTH_ISSUER (expected iss claim; default https://auth.shnt.top)
	AuthPortalURL  string // AUTH_PORTAL_URL (returned by /v1/auth/config for the UI)
	PortalTenantID string // fallback when JWT has no bgp_tenant_id / tenants.bgp
	AuthRequired   bool   // AUTH_REQUIRED / EVOBGP_AUTH_REQUIRED (surfaced via /v1/auth/config)
	// AuditIngestSecret — AUTH_AUDIT_INGEST_SECRET for portal push (optional).
	AuditIngestSecret string
}

// New constructs Server and wiring for async jobs.
func New(opts Options) (*Server, error) {
	backend, reg, pool, err := BootstrapWorkers(context.Background(), opts)
	if err != nil {
		return nil, err
	}

	var priv ed25519.PrivateKey
	if strings.TrimSpace(opts.BundleSeedHex) != "" {
		seed, err := hex.DecodeString(strings.TrimSpace(opts.BundleSeedHex))
		if err != nil {
			return nil, err
		}
		if len(seed) != ed25519.SeedSize {
			return nil, errors.New("httpapi: BundleSeedHex must decode to 32 bytes")
		}
		priv = ed25519.NewKeyFromSeed(seed)
	} else {
		_, priv, _ = ed25519.GenerateKey(rand.Reader)
	}

	resolver, err := newAPIKeyResolver(opts.APIKeys, backend)
	if err != nil {
		return nil, err
	}
	fwResolver, err := newFirewallTokenResolver(backend)
	if err != nil {
		return nil, err
	}
	var pgMon *pgmonitor.Service
	var maintCfg *maintenance.ConfigProvider
	var maintStats *maintenance.DBStatsProvider
	if pool != nil {
		pgMon = pgmonitor.NewService(pool)
		maintCfg = maintenance.NewConfigProvider(backend)
		_ = maintCfg.Reload(context.Background())
		maintStats = maintenance.NewDBStatsProvider(pgMon)
	}
	s := &Server{
		store:                   backend,
		pgPool:                  pool,
		pgMonitor:               pgMon,
		maintConfig:             maintCfg,
		maintStats:              maintStats,
		jobs:                    reg,
		bundlePriv:              priv,
		bundleCache:             newNodeBundleCache(nodeBundleCacheMaxEntries, nodeBundleCacheMaxBytes),
		keyResolver:             resolver,
		firewallResolver:        fwResolver,
		bundleSeedHex:           strings.TrimSpace(opts.BundleSeedHex),
		corsOrigins:             parseCORSOrigins(opts.CORSAllowedOrigins),
		cdnHTTP:                 NewCDNHTTPClient(),
		runtimeLogs:             runtimelogs.NewService(runtimelogs.ConfigFromEnv()),
		runtimeLogsPolicyTenant: strings.TrimSpace(opts.RuntimeLogsPolicyTenant),
		jwtSecret:               strings.TrimSpace(opts.JWTSecret),
		authIssuer:              strings.TrimSpace(opts.AuthIssuer),
		authPortalURL:           strings.TrimSpace(opts.AuthPortalURL),
		portalTenantID:          strings.TrimSpace(opts.PortalTenantID),
		authRequired:            opts.AuthRequired,
		auditIngestSecret:       strings.TrimSpace(opts.AuditIngestSecret),
	}
	if s.authIssuer == "" {
		s.authIssuer = "https://auth.shnt.top"
	}
	s.initAuditPusher(s.authPortalURL, s.auditIngestSecret)
	s.mux = http.NewServeMux()
	s.registerRoutes()
	return s, nil
}

// Close releases database resources.
func (s *Server) Close() {
	if s == nil {
		return
	}
	s.backgroundMu.Lock()
	if s.backgroundClosed {
		s.backgroundMu.Unlock()
		return
	}
	s.backgroundClosed = true
	cancel := s.backgroundCancel
	collector := s.telemetryCollector
	s.backgroundMu.Unlock()
	if cancel != nil {
		cancel()
	}
	if collector != nil {
		_ = collector.Close(context.Background())
	}
	// Do not close the pool while a worker is still persisting its final state.
	if s.jobs != nil {
		s.jobs.RequestCancelAll()
		_ = s.jobs.Drain(context.Background())
	}
	if s.pgPool != nil {
		s.pgPool.Close()
	}
}

// Store exposes the backing store (for operators / tests).
func (s *Server) Store() store.Backend { return s.store }

// Jobs exposes the in-process async job registry (for scheduler / evobgp-all).
func (s *Server) Jobs() *jobs.Registry { return s.jobs }

// StartBackground starts shared schedulers and telemetry polling once until ctx is cancelled.
func (s *Server) StartBackground(ctx context.Context) {
	if s == nil {
		return
	}
	s.backgroundMu.Lock()
	defer s.backgroundMu.Unlock()
	if s.backgroundClosed || s.backgroundStarted {
		return
	}
	s.backgroundStarted = true
	backgroundCtx, cancel := context.WithCancel(ctx)
	s.backgroundCancel = cancel
	if s.store != nil {
		s.telemetryCollector = telemetry.Start(backgroundCtx, s.store, telemetry.ConfigFromEnv())
	}
	if s.pgPool != nil {
		pgmonitor.StartScheduler(backgroundCtx, s.pgPool)
		if s.jobs != nil {
			jobs.StartDurableQueueLoop(backgroundCtx, s.jobs)
			jobs.StartDeploymentReconciliationLoop(backgroundCtx, s.jobs, s.store)
		}
	}
	if s != nil && s.maintConfig != nil && s.jobs != nil {
		maintenance.StartScheduler(backgroundCtx, s.maintConfig, func(policyID string, dryRun bool, idem string) {
			key := idem
			_, _, err := s.jobs.EnqueueContext(backgroundCtx, store.SystemJobTenantID, jobs.KindMaintenancePolicyRun, &key, nil, map[string]any{
				"policy_id": policyID,
				"dry_run":   dryRun,
				"trigger":   "scheduler",
			})
			if err != nil {
				logging.With("component", "maintenance.scheduler").Error("enqueue failed", "policy_id", policyID, "err", err)
			}
		}, 30*time.Second)
	}
	if s != nil && s.runtimeLogs != nil && s.store != nil {
		runtimelogs.StartAutoCleanupScheduler(backgroundCtx, runtimelogs.SchedulerDeps{
			Service:      s.runtimeLogs,
			Store:        s.store,
			PolicyTenant: s.runtimeLogsPolicyTenant,
		}, 30*time.Second)
	}
}