Log canceled requests into separate metrics

When a request get canceled we don't want to consider it an error

2 files changed, 41 insertions, 1 deletions

diff --git a/internal/sshd/connection.go b/internal/sshd/connection.go
index ebb4fe0..0295d8f 100644
--- a/internal/sshd/connection.go
+++ b/internal/sshd/connection.go
@@ -6,6 +6,8 @@ import (
 
 	"golang.org/x/crypto/ssh"
 	"golang.org/x/sync/semaphore"
+	grpccodes "google.golang.org/grpc/codes"
+	grpcstatus "google.golang.org/grpc/status"
 
 	"gitlab.com/gitlab-org/gitlab-shell/internal/config"
 	"gitlab.com/gitlab-org/gitlab-shell/internal/metrics"
@@ -85,7 +87,11 @@ func (c *connection) handle(ctx context.Context, chans <-chan ssh.NewChannel, ha
 			metrics.SliSshdSessionsTotal.Inc()
 			err := handler(ctx, channel, requests)
 			if err != nil {
-				metrics.SliSshdSessionsErrorsTotal.Inc()
+				if grpcstatus.Convert(err).Code() == grpccodes.Canceled {
+					metrics.SshdCanceledSessions.Inc()
+				} else {
+					metrics.SliSshdSessionsErrorsTotal.Inc()
+				}
 			}
 
 			ctxlog.Info("connection: handle: done")
diff --git a/internal/sshd/connection_test.go b/internal/sshd/connection_test.go
index d3a2287..f792300 100644
--- a/internal/sshd/connection_test.go
+++ b/internal/sshd/connection_test.go
@@ -7,10 +7,14 @@ import (
 	"testing"
 	"time"
 
+	"github.com/prometheus/client_golang/prometheus/testutil"
 	"github.com/stretchr/testify/require"
 	"golang.org/x/crypto/ssh"
+	grpccodes "google.golang.org/grpc/codes"
+	grpcstatus "google.golang.org/grpc/status"
 
 	"gitlab.com/gitlab-org/gitlab-shell/internal/config"
+	"gitlab.com/gitlab-org/gitlab-shell/internal/metrics"
 )
 
 type rejectCall struct {
@@ -189,3 +193,33 @@ func TestClientAliveInterval(t *testing.T) {
 
 	require.Eventually(t, func() bool { return KeepAliveMsg == f.SentRequestName() }, time.Second, time.Millisecond)
 }
+
+func TestSessionsMetrics(t *testing.T) {
+	// Unfortunately, there is no working way to reset Counter (not CounterVec)
+	// https://pkg.go.dev/github.com/prometheus/client_golang/prometheus#pkg-index
+	initialSessionsTotal := testutil.ToFloat64(metrics.SliSshdSessionsTotal)
+	initialSessionsErrorTotal := testutil.ToFloat64(metrics.SliSshdSessionsErrorsTotal)
+	initialCanceledSessions := testutil.ToFloat64(metrics.SshdCanceledSessions)
+
+	newChannel := &fakeNewChannel{channelType: "session"}
+
+	conn, chans := setup(1, newChannel)
+	conn.handle(context.Background(), chans, func(context.Context, ssh.Channel, <-chan *ssh.Request) error {
+		close(chans)
+		return errors.New("custom error")
+	})
+
+	require.InDelta(t, initialSessionsTotal+1, testutil.ToFloat64(metrics.SliSshdSessionsTotal), 0.1)
+	require.InDelta(t, initialSessionsErrorTotal+1, testutil.ToFloat64(metrics.SliSshdSessionsErrorsTotal), 0.1)
+	require.InDelta(t, initialCanceledSessions, testutil.ToFloat64(metrics.SshdCanceledSessions), 0.1)
+
+	conn, chans = setup(1, newChannel)
+	conn.handle(context.Background(), chans, func(context.Context, ssh.Channel, <-chan *ssh.Request) error {
+		close(chans)
+		return grpcstatus.Error(grpccodes.Canceled, "error")
+	})
+
+	require.InDelta(t, initialSessionsTotal+2, testutil.ToFloat64(metrics.SliSshdSessionsTotal), 0.1)
+	require.InDelta(t, initialSessionsErrorTotal+1, testutil.ToFloat64(metrics.SliSshdSessionsErrorsTotal), 0.1)
+	require.InDelta(t, initialCanceledSessions+1, testutil.ToFloat64(metrics.SshdCanceledSessions), 0.1)
+}