feat: Set default agent timeout to ~72h

mtojek · mtojek · commit aa3744db26f1 · 2023-06-16T12:03:45.000+02:00
diff --git a/agent/agentssh/agentssh.go b/agent/agentssh/agentssh.go
@@ -105,7 +105,7 @@ func NewServer(ctx context.Context, logger slog.Logger, prometheusRegistry *prom
 		metrics: metrics,
 	}
 
-	s.srv = &ssh.Server{
+	srv := &ssh.Server{
 		ChannelHandlers: map[string]ssh.ChannelHandler{
 			"direct-tcpip":                   ssh.DirectTCPIPHandler,
 			"direct-streamlocal@openssh.com": directStreamLocalHandler,
@@ -152,6 +152,12 @@ func NewServer(ctx context.Context, logger slog.Logger, prometheusRegistry *prom
 		MaxTimeout: maxTimeout,
 	}
 
+	if maxTimeout >= 3*time.Minute {
+		srv.ClientAliveInterval = maxTimeout / 3
+		srv.ClientAliveCountMax = 3
+	}
+
+	s.srv = srv
 	return s, nil
 }
 
diff --git a/agent/agentssh/agentssh_internal_test.go b/agent/agentssh/agentssh_internal_test.go
@@ -191,3 +191,7 @@ func (testSSHContext) Permissions() *gliderssh.Permissions {
 func (testSSHContext) SetValue(_, _ interface{}) {
 	panic("not implemented")
 }
+
+func (testSSHContext) KeepAlive() *gliderssh.SessionKeepAlive {
+	panic("not implemented")
+}
diff --git a/cli/agent.go b/cli/agent.go
@@ -323,8 +323,9 @@ func (r *RootCmd) workspaceAgent() *clibase.Cmd {
 			Value:       clibase.BoolOf(&noReap),
 		},
 		{
-			Flag:        "ssh-max-timeout",
-			Default:     "0",
+			Flag: "ssh-max-timeout",
+			// tcpip.KeepaliveIdleOption = 72h (forwardTCPSockOpts() in tailnet/conn.go)
+			Default:     fmt.Sprintf("%v", 72*time.Hour-time.Minute),
 			Env:         "CODER_AGENT_SSH_MAX_TIMEOUT",
 			Description: "Specify the max timeout for a SSH connection.",
 			Value:       clibase.DurationOf(&sshMaxTimeout),
diff --git a/cli/testdata/coder_agent_--help.golden b/cli/testdata/coder_agent_--help.golden
@@ -30,7 +30,7 @@ Starts the Coder workspace agent.
       --prometheus-address string, $CODER_AGENT_PROMETHEUS_ADDRESS (default: 127.0.0.1:2112)
           The bind address to serve Prometheus metrics.
 
-      --ssh-max-timeout duration, $CODER_AGENT_SSH_MAX_TIMEOUT (default: 0)
+      --ssh-max-timeout duration, $CODER_AGENT_SSH_MAX_TIMEOUT (default: 71h59m0s)
           Specify the max timeout for a SSH connection.
 
       --tailnet-listen-port int, $CODER_AGENT_TAILNET_LISTEN_PORT (default: 0)

Original file line number	Diff line number	Diff line change
`@@ -105,7 +105,7 @@ func NewServer(ctx context.Context, logger slog.Logger, prometheusRegistry *prom`
`105`	`105`	`metrics: metrics,`
`106`	`106`	`}`
`107`	`107`
`108`		`- s.srv = &ssh.Server{`
	`108`	`+ srv := &ssh.Server{`
`109`	`109`	`ChannelHandlers: map[string]ssh.ChannelHandler{`
`110`	`110`	`"direct-tcpip": ssh.DirectTCPIPHandler,`
`111`	`111`	`"direct-streamlocal@openssh.com": directStreamLocalHandler,`
`@@ -152,6 +152,12 @@ func NewServer(ctx context.Context, logger slog.Logger, prometheusRegistry *prom`
`152`	`152`	`MaxTimeout: maxTimeout,`
`153`	`153`	`}`
`154`	`154`
	`155`	`+ if maxTimeout >= 3*time.Minute {`
	`156`	`+ srv.ClientAliveInterval = maxTimeout / 3`
	`157`	`+ srv.ClientAliveCountMax = 3`
	`158`	`+ }`
	`159`	`+`
	`160`	`+ s.srv = srv`
`155`	`161`	`return s, nil`
`156`	`162`	`}`
`157`	`163`
Original file line number	Diff line number	Diff line change
`@@ -191,3 +191,7 @@ func (testSSHContext) Permissions() *gliderssh.Permissions {`
`191`	`191`	`func (testSSHContext) SetValue(_, _ interface{}) {`
`192`	`192`	`panic("not implemented")`
`193`	`193`	`}`
	`194`	`+`
	`195`	`+func (testSSHContext) KeepAlive() *gliderssh.SessionKeepAlive {`
	`196`	`+ panic("not implemented")`
	`197`	`+}`