fix(tailnet): track agent names for http debug (#8744)

coadler · web-flow · commit 6b92abebb99f · 2023-07-26T18:44:10.000Z
diff --git a/enterprise/tailnet/coordinator.go b/enterprise/tailnet/coordinator.go
@@ -158,7 +158,7 @@ func (c *haCoordinator) ServeClient(conn net.Conn, id, agentID uuid.UUID) error
 	defer cancel()
 	logger := c.clientLogger(id, agentID)
 
-	tc := agpl.NewTrackedConn(ctx, cancel, conn, id, logger, 0)
+	tc := agpl.NewTrackedConn(ctx, cancel, conn, id, logger, id.String(), 0)
 	defer tc.Close()
 
 	c.addClient(id, tc)
@@ -301,7 +301,7 @@ func (c *haCoordinator) ServeAgent(conn net.Conn, id uuid.UUID, name string) err
 	}
 	// This uniquely identifies a connection that belongs to this goroutine.
 	unique := uuid.New()
-	tc := agpl.NewTrackedConn(ctx, cancel, conn, unique, logger, overwrites)
+	tc := agpl.NewTrackedConn(ctx, cancel, conn, unique, logger, name, overwrites)
 
 	// Publish all nodes on this instance that want to connect to this agent.
 	nodes := c.nodesSubscribedToAgent(id)
diff --git a/enterprise/tailnet/pgcoord.go b/enterprise/tailnet/pgcoord.go
@@ -131,12 +131,6 @@ func (c *pgCoord) ServeMultiAgent(id uuid.UUID) agpl.MultiAgentConn {
 	panic("not implemented") // TODO: Implement
 }
 
-func (*pgCoord) ServeHTTPDebug(w http.ResponseWriter, _ *http.Request) {
-	// TODO(spikecurtis) I'd like to hold off implementing this until after the rest of this is code reviewed.
-	w.WriteHeader(http.StatusOK)
-	_, _ = w.Write([]byte("Coder Enterprise PostgreSQL distributed tailnet coordinator"))
-}
-
 func (c *pgCoord) Node(id uuid.UUID) *agpl.Node {
 	// In production, we only ever get this request for an agent.
 	// We're going to directly query the database, since we would only have the agent mapping stored locally if we had
@@ -167,7 +161,7 @@ func (c *pgCoord) ServeClient(conn net.Conn, id uuid.UUID, agent uuid.UUID) erro
 				slog.Error(err))
 		}
 	}()
-	cIO := newConnIO(c.ctx, c.logger, c.bindings, conn, id, agent)
+	cIO := newConnIO(c.ctx, c.logger, c.bindings, conn, id, agent, id.String())
 	if err := sendCtx(c.ctx, c.newConnections, cIO); err != nil {
 		// can only be a context error, no need to log here.
 		return err
@@ -186,7 +180,7 @@ func (c *pgCoord) ServeAgent(conn net.Conn, id uuid.UUID, name string) error {
 		}
 	}()
 	logger := c.logger.With(slog.F("name", name))
-	cIO := newConnIO(c.ctx, logger, c.bindings, conn, uuid.Nil, id)
+	cIO := newConnIO(c.ctx, logger, c.bindings, conn, uuid.Nil, id, name)
 	if err := sendCtx(c.ctx, c.newConnections, cIO); err != nil {
 		// can only be a context error, no need to log here.
 		return err
@@ -217,8 +211,12 @@ type connIO struct {
 	bindings chan<- binding
 }
 
-func newConnIO(
-	pCtx context.Context, logger slog.Logger, bindings chan<- binding, conn net.Conn, client, agent uuid.UUID,
+func newConnIO(pCtx context.Context,
+	logger slog.Logger,
+	bindings chan<- binding,
+	conn net.Conn,
+	client, agent uuid.UUID,
+	name string,
 ) *connIO {
 	ctx, cancel := context.WithCancel(pCtx)
 	id := agent
@@ -235,7 +233,7 @@ func newConnIO(
 		client:   client,
 		agent:    agent,
 		decoder:  json.NewDecoder(conn),
-		updates:  agpl.NewTrackedConn(ctx, cancel, conn, id, logger, 0),
+		updates:  agpl.NewTrackedConn(ctx, cancel, conn, id, logger, name, 0),
 		bindings: bindings,
 	}
 	go c.recvLoop()
@@ -1291,3 +1289,9 @@ func (h *heartbeats) cleanup() {
 	}
 	h.logger.Debug(h.ctx, "cleaned up old coordinators")
 }
+
+func (*pgCoord) ServeHTTPDebug(w http.ResponseWriter, _ *http.Request) {
+	// TODO(spikecurtis) I'd like to hold off implementing this until after the rest of this is code reviewed.
+	w.WriteHeader(http.StatusOK)
+	_, _ = w.Write([]byte("Coder Enterprise PostgreSQL distributed tailnet coordinator"))
+}
diff --git a/go.mod b/go.mod
@@ -91,6 +91,7 @@ require (
 	github.com/coreos/go-systemd v0.0.0-20191104093116-d3cd4ed1dbcf
 	github.com/creack/pty v1.1.18
 	github.com/dave/dst v0.27.2
+	github.com/davecgh/go-spew v1.1.1
 	github.com/elastic/go-sysinfo v1.11.0
 	github.com/fatih/color v1.15.0
 	github.com/fatih/structs v1.1.0
@@ -187,6 +188,7 @@ require (
 	nhooyr.io/websocket v1.8.7
 	storj.io/drpc v0.0.33-0.20230420154621-9716137f6037
 	tailscale.com v1.32.3
+	golang.org/x/net v0.12.0
 )
 
 require (
@@ -222,7 +224,6 @@ require (
 	github.com/containerd/console v1.0.3 // indirect
 	github.com/containerd/continuity v0.3.0 // indirect
 	github.com/coreos/go-iptables v0.6.0 // indirect
-	github.com/davecgh/go-spew v1.1.1 // indirect
 	github.com/dlclark/regexp2 v1.10.0 // indirect
 	github.com/docker/cli v20.10.17+incompatible // indirect
 	github.com/docker/docker v23.0.3+incompatible // indirect
@@ -355,7 +356,6 @@ require (
 	go.opentelemetry.io/otel/metric v1.16.0 // indirect
 	go.opentelemetry.io/proto/otlp v0.19.0 // indirect
 	go4.org/mem v0.0.0-20210711025021-927187094b94 // indirect
-	golang.org/x/net v0.12.0
 	golang.org/x/text v0.11.0 // indirect
 	golang.org/x/time v0.3.0 // indirect
 	golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2 // indirect
diff --git a/tailnet/coordinator.go b/tailnet/coordinator.go
@@ -262,7 +262,7 @@ func (c *coordinator) ServeClient(conn net.Conn, id, agentID uuid.UUID) error {
 	logger := c.core.clientLogger(id, agentID)
 	logger.Debug(ctx, "coordinating client")
 
-	tc := NewTrackedConn(ctx, cancel, conn, id, logger, 0)
+	tc := NewTrackedConn(ctx, cancel, conn, id, logger, id.String(), 0)
 	defer tc.Close()
 
 	c.core.addClient(id, tc)
@@ -507,7 +507,7 @@ func (c *core) initAndTrackAgent(ctx context.Context, cancel func(), conn net.Co
 		overwrites = oldAgentSocket.Overwrites() + 1
 		_ = oldAgentSocket.Close()
 	}
-	tc := NewTrackedConn(ctx, cancel, conn, unique, logger, overwrites)
+	tc := NewTrackedConn(ctx, cancel, conn, unique, logger, name, overwrites)
 	c.agentNameCache.Add(id, name)
 
 	sockets, ok := c.agentToConnectionSockets[id]
diff --git a/tailnet/trackedconn.go b/tailnet/trackedconn.go
@@ -35,7 +35,7 @@ type TrackedConn struct {
 	overwrites int64
 }
 
-func NewTrackedConn(ctx context.Context, cancel func(), conn net.Conn, id uuid.UUID, logger slog.Logger, overwrites int64) *TrackedConn {
+func NewTrackedConn(ctx context.Context, cancel func(), conn net.Conn, id uuid.UUID, logger slog.Logger, name string, overwrites int64) *TrackedConn {
 	// buffer updates so they don't block, since we hold the
 	// coordinator mutex while queuing.  Node updates don't
 	// come quickly, so 512 should be plenty for all but
@@ -51,6 +51,7 @@ func NewTrackedConn(ctx context.Context, cancel func(), conn net.Conn, id uuid.U
 		id:         id,
 		start:      now,
 		lastWrite:  now,
+		name:       name,
 		overwrites: overwrites,
 	}
 }

Original file line number	Diff line number	Diff line change
`@@ -35,7 +35,7 @@ type TrackedConn struct {`
`35`	`35`	`overwrites int64`
`36`	`36`	`}`
`37`	`37`
`38`		`-func NewTrackedConn(ctx context.Context, cancel func(), conn net.Conn, id uuid.UUID, logger slog.Logger, overwrites int64) *TrackedConn {`
	`38`	`+func NewTrackedConn(ctx context.Context, cancel func(), conn net.Conn, id uuid.UUID, logger slog.Logger, name string, overwrites int64) *TrackedConn {`
`39`	`39`	`// buffer updates so they don't block, since we hold the`
`40`	`40`	`// coordinator mutex while queuing. Node updates don't`
`41`	`41`	`// come quickly, so 512 should be plenty for all but`
`@@ -51,6 +51,7 @@ func NewTrackedConn(ctx context.Context, cancel func(), conn net.Conn, id uuid.U`
`51`	`51`	`id: id,`
`52`	`52`	`start: now,`
`53`	`53`	`lastWrite: now,`
	`54`	`+ name: name,`
`54`	`55`	`overwrites: overwrites,`
`55`	`56`	`}`
`56`	`57`	`}`