Merge branch 'id-improve-errors-metrics' into 'main'

Exclude authentication errors from error rate

See merge request gitlab-org/gitlab-shell!611

2 files changed, 14 insertions, 15 deletions

diff --git a/internal/sshd/connection.go b/internal/sshd/connection.go
index 1312833..278bb37 100644
--- a/internal/sshd/connection.go
+++ b/internal/sshd/connection.go
@@ -2,6 +2,7 @@ package sshd
 
 import (
 	"context"
+	"time"
 
 	"golang.org/x/crypto/ssh"
 	"golang.org/x/sync/semaphore"
@@ -49,6 +50,10 @@ func (c *connection) handle(ctx context.Context, chans <-chan ssh.NewChannel, ha
 		}
 
 		go func() {
+			defer func(started time.Time) {
+				metrics.SshdSessionDuration.Observe(time.Since(started).Seconds())
+			}(time.Now())
+
 			defer c.concurrentSessions.Release(1)
 
 			// Prevent a panic in a single session from taking out the whole server
diff --git a/internal/sshd/sshd.go b/internal/sshd/sshd.go
index 49b8ab9..d6ea6c6 100644
--- a/internal/sshd/sshd.go
+++ b/internal/sshd/sshd.go
@@ -146,19 +146,8 @@ func (s *Server) getStatus() status {
 }
 
 func (s *Server) handleConn(ctx context.Context, nconn net.Conn) {
-	success := false
-
 	metrics.SshdConnectionsInFlight.Inc()
-	started := time.Now()
-	defer func() {
-		metrics.SshdConnectionsInFlight.Dec()
-		metrics.SshdSessionDuration.Observe(time.Since(started).Seconds())
-
-		metrics.SliSshdSessionsTotal.Inc()
-		if !success {
-			metrics.SliSshdSessionsErrorsTotal.Inc()
-		}
-	}()
+	defer metrics.SshdConnectionsInFlight.Dec()
 
 	remoteAddr := nconn.RemoteAddr().String()
 
@@ -174,6 +163,8 @@ func (s *Server) handleConn(ctx context.Context, nconn net.Conn) {
 	defer func() {
 		if err := recover(); err != nil {
 			ctxlog.Warn("panic handling session")
+
+			metrics.SliSshdSessionsErrorsTotal.Inc()
 		}
 	}()
 
@@ -181,11 +172,12 @@ func (s *Server) handleConn(ctx context.Context, nconn net.Conn) {
 
 	sconn, chans, reqs, err := ssh.NewServerConn(nconn, s.serverConfig.get(ctx))
 	if err != nil {
-		ctxlog.WithError(err).Error("server: handleConn: failed to initialize SSH connection")
+		ctxlog.WithError(err).Warn("server: handleConn: failed to initialize SSH connection")
 		return
 	}
 	go ssh.DiscardRequests(reqs)
 
+	started := time.Now()
 	var establishSessionDuration float64
 	conn := newConnection(s.Config.Server.ConcurrentSessionsLimit, remoteAddr)
 	conn.handle(ctx, chans, func(ctx context.Context, channel ssh.Channel, requests <-chan *ssh.Request) {
@@ -199,9 +191,11 @@ func (s *Server) handleConn(ctx context.Context, nconn net.Conn) {
 			remoteAddr:  remoteAddr,
 		}
 
+		metrics.SliSshdSessionsTotal.Inc()
 		session.handle(ctx, requests)
-
-		success = session.success
+		if !session.success {
+			metrics.SliSshdSessionsErrorsTotal.Inc()
+		}
 	})
 
 	ctxlog.WithFields(log.Fields{