log error with usage

livekit · Jan 8, 2024 · 04ad939 · 04ad939
1 parent c4ce014
commit 04ad939
Show file tree

Hide file tree

Showing 3 changed files with 8 additions and 6 deletions.
diff --git a/pkg/errors/errors.go b/pkg/errors/errors.go
@@ -27,7 +27,6 @@ import (
 var (
 	ErrNoConfig                   = psrpc.NewErrorf(psrpc.Internal, "missing config")
 	ErrGhostPadFailed             = psrpc.NewErrorf(psrpc.Internal, "failed to add ghost pad to bin")
-	ErrStreamAlreadyExists        = psrpc.NewErrorf(psrpc.AlreadyExists, "stream already exists")
 	ErrBinAlreadyAdded            = psrpc.NewErrorf(psrpc.Internal, "bin already added to pipeline")
 	ErrWrongHierarchy             = psrpc.NewErrorf(psrpc.Internal, "pipeline can contain bins or elements, not both")
 	ErrNonStreamingPipeline       = psrpc.NewErrorf(psrpc.InvalidArgument, "UpdateStream called on non-streaming egress")
@@ -38,6 +37,7 @@ var (
 	ErrSubscriptionFailed         = psrpc.NewErrorf(psrpc.Internal, "failed to subscribe to track")
 	ErrPipelineFrozen             = psrpc.NewErrorf(psrpc.Internal, "pipeline frozen")
 	ErrSinkNotFound               = psrpc.NewErrorf(psrpc.Internal, "sink not found")
+	ErrCPUExhausted               = psrpc.NewErrorf(psrpc.Unavailable, "CPU exhausted")
 )
 
 func New(err string) error {

diff --git a/pkg/service/service.go b/pkg/service/service.go
@@ -30,6 +30,7 @@ import (
 	"google.golang.org/grpc"
 
 	"github.com/livekit/egress/pkg/config"
+	"github.com/livekit/egress/pkg/errors"
 	"github.com/livekit/egress/pkg/ipc"
 	"github.com/livekit/egress/pkg/stats"
 	"github.com/livekit/egress/version"
@@ -183,12 +184,13 @@ func (s *Service) KillAll() {
 	}
 }
 
-func (s *Service) killProcess(egressID string) {
+func (s *Service) killProcess(egressID string, maxUsage float64) {
 	s.mu.RLock()
 	defer s.mu.RUnlock()
 
 	if h, ok := s.activeHandlers[egressID]; ok {
-		h.info.Error = "CPU exhausted"
+		logger.Errorw("killing egress", errors.ErrCPUExhausted, "egressID", egressID, "usage", maxUsage)
+		h.info.Error = errors.ErrCPUExhausted.Error()
 		h.kill()
 	}
 }

diff --git a/pkg/stats/monitor.go b/pkg/stats/monitor.go
@@ -44,7 +44,7 @@ type Monitor struct {
 	mu              sync.Mutex
 	highCPUDuration int
 	killThreshold   float64
-	killProcess     func(string)
+	killProcess     func(string, float64)
 	pending         map[string]*processStats
 	procStats       map[int]*processStats
 }
@@ -85,7 +85,7 @@ func (m *Monitor) Start(
 	conf *config.ServiceConfig,
 	isIdle func() float64,
 	canAcceptRequest func() float64,
-	killProcess func(string),
+	killProcess func(string, float64),
 ) error {
 	m.killProcess = killProcess
 
@@ -241,7 +241,7 @@ func (m *Monitor) updateEgressStats(idle float64, usage map[int]float64) {
 			if m.highCPUDuration < minKillDuration {
 				return
 			}
-			m.killProcess(maxEgress)
+			m.killProcess(maxEgress, maxUsage)
 		}
 	}