vpp-maglev/internal/grpcapi/server.go

// Copyright (c) 2026, Pim van Pelt <pim@ipng.ch>

package grpcapi

import (
	"context"
	"log/slog"
	"net"

	"google.golang.org/grpc/codes"
	"google.golang.org/grpc/status"

	"git.ipng.ch/ipng/vpp-maglev/internal/checker"
	"git.ipng.ch/ipng/vpp-maglev/internal/config"
	"git.ipng.ch/ipng/vpp-maglev/internal/health"
)

// Server implements the MaglevServer gRPC interface.
type Server struct {
	UnimplementedMaglevServer
	ctx        context.Context
	checker    *checker.Checker
	logs       *LogBroadcaster
	configPath string
}

// NewServer creates a Server backed by the given Checker. logs may be nil, in
// which case log events are never sent to WatchEvents streams. configPath is
// used by CheckConfig to reload and validate the configuration file on demand.
// The provided context controls the lifetime of streaming RPCs: cancelling it
// closes all active WatchEvents streams so that grpc.Server.GracefulStop can
// complete.
func NewServer(ctx context.Context, c *checker.Checker, logs *LogBroadcaster, configPath string) *Server {
	return &Server{ctx: ctx, checker: c, logs: logs, configPath: configPath}
}

// ListFrontends returns the names of all configured frontends.
func (s *Server) ListFrontends(_ context.Context, _ *ListFrontendsRequest) (*ListFrontendsResponse, error) {
	return &ListFrontendsResponse{FrontendNames: s.checker.ListFrontends()}, nil
}

// GetFrontend returns configuration details for a single frontend.
func (s *Server) GetFrontend(_ context.Context, req *GetFrontendRequest) (*FrontendInfo, error) {
	fe, ok := s.checker.GetFrontend(req.Name)
	if !ok {
		return nil, status.Errorf(codes.NotFound, "frontend %q not found", req.Name)
	}
	return frontendToProto(req.Name, fe), nil
}

// ListBackends returns the names of all active backends.
func (s *Server) ListBackends(_ context.Context, _ *ListBackendsRequest) (*ListBackendsResponse, error) {
	return &ListBackendsResponse{BackendNames: s.checker.ListBackends()}, nil
}

// GetBackend returns health state for a backend by name.
func (s *Server) GetBackend(_ context.Context, req *GetBackendRequest) (*BackendInfo, error) {
	b, ok := s.checker.GetBackend(req.Name)
	if !ok {
		return nil, status.Errorf(codes.NotFound, "backend %q not found", req.Name)
	}
	return backendToProto(b), nil
}

// PauseBackend pauses health checking for a backend by name.
func (s *Server) PauseBackend(_ context.Context, req *BackendRequest) (*BackendInfo, error) {
	b, ok := s.checker.PauseBackend(req.Name)
	if !ok {
		return nil, status.Errorf(codes.NotFound, "backend %q not found", req.Name)
	}
	return backendToProto(b), nil
}

// ResumeBackend resumes health checking for a backend by name.
func (s *Server) ResumeBackend(_ context.Context, req *BackendRequest) (*BackendInfo, error) {
	b, ok := s.checker.ResumeBackend(req.Name)
	if !ok {
		return nil, status.Errorf(codes.NotFound, "backend %q not found", req.Name)
	}
	return backendToProto(b), nil
}

// EnableBackend re-enables a previously disabled backend.
func (s *Server) EnableBackend(_ context.Context, req *BackendRequest) (*BackendInfo, error) {
	b, ok := s.checker.EnableBackend(req.Name)
	if !ok {
		return nil, status.Errorf(codes.NotFound, "backend %q not found", req.Name)
	}
	return backendToProto(b), nil
}

// DisableBackend disables a backend, stopping its probe goroutine.
func (s *Server) DisableBackend(_ context.Context, req *BackendRequest) (*BackendInfo, error) {
	b, ok := s.checker.DisableBackend(req.Name)
	if !ok {
		return nil, status.Errorf(codes.NotFound, "backend %q not found", req.Name)
	}
	return backendToProto(b), nil
}

// SetFrontendPoolBackendWeight updates the weight of a backend in a pool.
func (s *Server) SetFrontendPoolBackendWeight(_ context.Context, req *SetWeightRequest) (*FrontendInfo, error) {
	if req.Weight < 0 || req.Weight > 100 {
		return nil, status.Errorf(codes.InvalidArgument, "weight %d out of range [0, 100]", req.Weight)
	}
	fe, err := s.checker.SetFrontendPoolBackendWeight(req.Frontend, req.Pool, req.Backend, int(req.Weight))
	if err != nil {
		return nil, status.Errorf(codes.NotFound, "%v", err)
	}
	return frontendToProto(req.Frontend, fe), nil
}

// ListHealthChecks returns the names of all configured health checks.
func (s *Server) ListHealthChecks(_ context.Context, _ *ListHealthChecksRequest) (*ListHealthChecksResponse, error) {
	return &ListHealthChecksResponse{Names: s.checker.ListHealthChecks()}, nil
}

// GetHealthCheck returns the full configuration for a health check by name.
func (s *Server) GetHealthCheck(_ context.Context, req *GetHealthCheckRequest) (*HealthCheckInfo, error) {
	hc, ok := s.checker.GetHealthCheck(req.Name)
	if !ok {
		return nil, status.Errorf(codes.NotFound, "healthcheck %q not found", req.Name)
	}
	return healthCheckToProto(req.Name, hc), nil
}

// WatchEvents streams events to the client. On connect, the current state of
// all backends is sent as synthetic BackendEvents. Afterwards, live events are
// forwarded based on the filter flags in req. An unset (nil) flag defaults to
// true (subscribe). An empty log_level defaults to "info".
func (s *Server) WatchEvents(req *WatchRequest, stream Maglev_WatchEventsServer) error {
	wantLog := req.Log == nil || *req.Log
	wantBackend := req.Backend == nil || *req.Backend
	wantFrontend := req.Frontend == nil || *req.Frontend
	_ = wantFrontend // no frontend events emitted yet

	logLevel := slog.LevelInfo
	if req.LogLevel != "" {
		if err := logLevel.UnmarshalText([]byte(req.LogLevel)); err != nil {
			return status.Errorf(codes.InvalidArgument, "invalid log_level %q: must be debug, info, warn, or error", req.LogLevel)
		}
	}

	// Subscribe to log events (nil channel blocks forever when not wanted).
	var logCh <-chan *LogEvent
	if wantLog && s.logs != nil {
		var unsub func()
		logCh, unsub = s.logs.Subscribe(logLevel)
		defer unsub()
	}

	// Subscribe to backend events; send initial state snapshot first.
	var backendCh <-chan checker.Event
	if wantBackend {
		for _, name := range s.checker.ListBackends() {
			snap, ok := s.checker.GetBackend(name)
			if !ok {
				continue
			}
			ev := &Event{Event: &Event_Backend{Backend: &BackendEvent{
				BackendName: name,
				Transition: &TransitionRecord{
					From:     snap.Health.State.String(),
					To:       snap.Health.State.String(),
					AtUnixNs: 0,
				},
			}}}
			if err := stream.Send(ev); err != nil {
				return err
			}
		}
		var unsub func()
		backendCh, unsub = s.checker.Subscribe()
		defer unsub()
	}

	for {
		select {
		case <-s.ctx.Done():
			return status.Error(codes.Unavailable, "server shutting down")
		case <-stream.Context().Done():
			return nil
		case le, ok := <-logCh:
			if !ok {
				return nil
			}
			if err := stream.Send(&Event{Event: &Event_Log{Log: le}}); err != nil {
				return err
			}
		case e, ok := <-backendCh:
			if !ok {
				return nil
			}
			if err := stream.Send(&Event{Event: &Event_Backend{Backend: &BackendEvent{
				BackendName: e.BackendName,
				Transition:  transitionToProto(e.Transition),
			}}}); err != nil {
				return err
			}
		}
	}
}

// CheckConfig reads and validates the configuration file, returning a
// structured result that distinguishes YAML parse errors from semantic errors.
func (s *Server) CheckConfig(_ context.Context, _ *CheckConfigRequest) (*CheckConfigResponse, error) {
	slog.Info("config-check-start", "path", s.configPath)
	_, result := config.Check(s.configPath)
	resp := &CheckConfigResponse{
		Ok:            result.OK(),
		ParseError:    result.ParseError,
		SemanticError: result.SemanticError,
	}
	if result.OK() {
		slog.Info("config-check-done", "result", "ok")
	} else if result.ParseError != "" {
		slog.Info("config-check-done", "result", "failed", "type", "parse", "err", result.ParseError)
	} else {
		slog.Info("config-check-done", "result", "failed", "type", "semantic", "err", result.SemanticError)
	}
	return resp, nil
}

// ReloadConfig checks the configuration file and, if valid, applies it to the
// running checker. This is the same code path used by SIGHUP.
func (s *Server) ReloadConfig(_ context.Context, _ *ReloadConfigRequest) (*ReloadConfigResponse, error) {
	return s.doReloadConfig(), nil
}

// TriggerReload performs a config check and reload. Intended for use by the
// SIGHUP handler so that signals and gRPC share the same code path.
func (s *Server) TriggerReload() {
	s.doReloadConfig()
}

func (s *Server) doReloadConfig() *ReloadConfigResponse {
	slog.Info("config-reload-start")
	newCfg, result := config.Check(s.configPath)
	if !result.OK() {
		if result.ParseError != "" {
			slog.Error("config-check-failed", "type", "parse", "err", result.ParseError)
		} else {
			slog.Error("config-check-failed", "type", "semantic", "err", result.SemanticError)
		}
		return &ReloadConfigResponse{
			ParseError:    result.ParseError,
			SemanticError: result.SemanticError,
		}
	}
	if err := s.checker.Reload(s.ctx, newCfg); err != nil {
		slog.Error("checker-reload-error", "err", err)
		return &ReloadConfigResponse{
			ReloadError: err.Error(),
		}
	}
	slog.Info("config-reload-done", "frontends", len(newCfg.Frontends))
	return &ReloadConfigResponse{Ok: true}
}

// ---- conversion helpers ----------------------------------------------------

func frontendToProto(name string, fe config.Frontend) *FrontendInfo {
	pools := make([]*PoolInfo, 0, len(fe.Pools))
	for _, p := range fe.Pools {
		pi := &PoolInfo{Name: p.Name}
		for bName, pb := range p.Backends {
			pi.Backends = append(pi.Backends, &PoolBackendInfo{
				Name:   bName,
				Weight: int32(pb.Weight),
			})
		}
		pools = append(pools, pi)
	}
	return &FrontendInfo{
		Name:        name,
		Address:     fe.Address.String(),
		Protocol:    fe.Protocol,
		Port:        uint32(fe.Port),
		Description: fe.Description,
		Pools:       pools,
	}
}

func backendToProto(snap checker.BackendSnapshot) *BackendInfo {
	info := &BackendInfo{
		Name:        snap.Health.Name,
		Address:     snap.Health.Address.String(),
		State:       snap.Health.State.String(),
		Enabled:     snap.Config.Enabled,
		Healthcheck: snap.Config.HealthCheck,
	}
	for _, t := range snap.Health.Transitions {
		info.Transitions = append(info.Transitions, transitionToProto(t))
	}
	return info
}

func healthCheckToProto(name string, hc config.HealthCheck) *HealthCheckInfo {
	info := &HealthCheckInfo{
		Name:           name,
		Type:           hc.Type,
		Port:           uint32(hc.Port),
		IntervalNs:     hc.Interval.Nanoseconds(),
		FastIntervalNs: hc.FastInterval.Nanoseconds(),
		DownIntervalNs: hc.DownInterval.Nanoseconds(),
		TimeoutNs:      hc.Timeout.Nanoseconds(),
		Rise:           int32(hc.Rise),
		Fall:           int32(hc.Fall),
	}
	if hc.ProbeIPv4Src != nil {
		info.ProbeIpv4Src = hc.ProbeIPv4Src.String()
	}
	if hc.ProbeIPv6Src != nil {
		info.ProbeIpv6Src = hc.ProbeIPv6Src.String()
	}
	if hc.HTTP != nil {
		re := ""
		if hc.HTTP.ResponseRegexp != nil {
			re = hc.HTTP.ResponseRegexp.String()
		}
		info.Http = &HTTPCheckParams{
			Path:               hc.HTTP.Path,
			Host:               hc.HTTP.Host,
			ResponseCodeMin:    int32(hc.HTTP.ResponseCodeMin),
			ResponseCodeMax:    int32(hc.HTTP.ResponseCodeMax),
			ResponseRegexp:     re,
			ServerName:         hc.HTTP.ServerName,
			InsecureSkipVerify: hc.HTTP.InsecureSkipVerify,
		}
	}
	if hc.TCP != nil {
		info.Tcp = &TCPCheckParams{
			Ssl:                hc.TCP.SSL,
			ServerName:         hc.TCP.ServerName,
			InsecureSkipVerify: hc.TCP.InsecureSkipVerify,
		}
	}
	return info
}

func transitionToProto(t health.Transition) *TransitionRecord {
	return &TransitionRecord{
		From:     t.From.String(),
		To:       t.To.String(),
		AtUnixNs: t.At.UnixNano(),
	}
}

// Ensure net.IP is imported (used via b.Address.String()).
var _ = net.IP{}