feat(rpc/node): readiness check (#3118)

Implements a readiness check on the node module. Question: I have set the permission level to public, but remember that we removed all public methods in favor of `read`. Is this fine or should this method also be `read`? ____________ Comes from a conversation with @joroshiba from Astria --------- Co-authored-by: ramin <[email protected]>
celestiaorg · Jan 29, 2024 · 9484e15 · 9484e15
1 parent 4850c90
commit 9484e15
Show file tree

Hide file tree

Showing 4 changed files with 21 additions and 1 deletion.
diff --git a/nodebuilder/module.go b/nodebuilder/module.go
@@ -50,14 +50,14 @@ func ConstructModule(tp node.Type, network p2p.Network, cfg *Config, store Store
 		state.ConstructModule(tp, &cfg.State, &cfg.Core),
 		modhead.ConstructModule[*header.ExtendedHeader](tp, &cfg.Header),
 		share.ConstructModule(tp, &cfg.Share),
-		rpc.ConstructModule(tp, &cfg.RPC),
 		gateway.ConstructModule(tp, &cfg.Gateway),
 		core.ConstructModule(tp, &cfg.Core),
 		das.ConstructModule(tp, &cfg.DASer),
 		fraud.ConstructModule(tp),
 		blob.ConstructModule(),
 		node.ConstructModule(tp),
 		prune.ConstructModule(tp),
+		rpc.ConstructModule(tp, &cfg.RPC),
 	)
 
 	return fx.Module(

diff --git a/nodebuilder/node/admin.go b/nodebuilder/node/admin.go
@@ -38,6 +38,14 @@ func (m *module) Info(context.Context) (Info, error) {
 	}, nil
 }
 
+func (m *module) Ready(context.Context) (bool, error) {
+	// Because the node uses FX to provide the RPC last, all services' lifecycles have been started by
+	// the point this endpoint is available. It is not 100% guaranteed at this point that all services
+	// are fully ready, but it is very high probability and all endpoints are available at this point
+	// regardless.
+	return true, nil
+}
+
 func (m *module) LogLevelSet(_ context.Context, name, level string) error {
 	return logging.SetLogLevel(name, level)
 }

diff --git a/nodebuilder/node/node.go b/nodebuilder/node/node.go
@@ -14,6 +14,9 @@ type Module interface {
 	// Info returns administrative information about the node.
 	Info(context.Context) (Info, error)
 
+	// Ready returns true once the node's RPC is ready to accept requests.
+	Ready(context.Context) (bool, error)
+
 	// LogLevelSet sets the given component log level to the given level.
 	LogLevelSet(ctx context.Context, name, level string) error
 
@@ -28,6 +31,7 @@ var _ Module = (*API)(nil)
 type API struct {
 	Internal struct {
 		Info        func(context.Context) (Info, error)                                `perm:"admin"`
+		Ready       func(context.Context) (bool, error)                                `perm:"read"`
 		LogLevelSet func(ctx context.Context, name, level string) error                `perm:"admin"`
 		AuthVerify  func(ctx context.Context, token string) ([]auth.Permission, error) `perm:"admin"`
 		AuthNew     func(ctx context.Context, perms []auth.Permission) (string, error) `perm:"admin"`
@@ -38,6 +42,10 @@ func (api *API) Info(ctx context.Context) (Info, error) {
 	return api.Internal.Info(ctx)
 }
 
+func (api *API) Ready(ctx context.Context) (bool, error) {
+	return api.Internal.Ready(ctx)
+}
+
 func (api *API) LogLevelSet(ctx context.Context, name, level string) error {
 	return api.Internal.LogLevelSet(ctx, name, level)
 }

diff --git a/nodebuilder/tests/api_test.go b/nodebuilder/tests/api_test.go
@@ -48,6 +48,10 @@ func TestNodeModule(t *testing.T) {
 	require.NoError(t, err)
 	require.Equal(t, info.APIVersion, node.APIVersion)
 
+	ready, err := client.Node.Ready(ctx)
+	require.NoError(t, err)
+	require.True(t, ready)
+
 	perms, err := client.Node.AuthVerify(ctx, jwt)
 	require.NoError(t, err)
 	require.Equal(t, perms, adminPerms)