core: Save app provisioning errors with context (#7070)

* fix(provisioning): remove app from apps map when its provision failed

* Clean up failed app provisioning with defer

* fix(provisioning): record apps that failed to provision with their error

* save the error when an app fails to initialize and return this error when this app is requested by a module

---------

Co-authored-by: Matt Holt <mholt@users.noreply.github.com>
Co-authored-by: WeidiDeng <weidi_deng@icloud.com>

Author: 3 people

caddy.go

@@ -81,7 +81,10 @@ type Config struct {
 	// associated value.
 	AppsRaw ModuleMap `json:"apps,omitempty" caddy:"namespace="`
 
-	apps         map[string]App
+	apps map[string]App
+
+	// failedApps is a map of apps that failed to provision with their underlying error.
+	failedApps   map[string]error
 	storage      certmagic.Storage
 	eventEmitter eventEmitter
 
@@ -522,6 +525,7 @@ func provisionContext(newCfg *Config, replaceAdminServer bool) (Context, error)
 
 	// prepare the new config for use
 	newCfg.apps = make(map[string]App)
+	newCfg.failedApps = make(map[string]error)
 
 	// set up global storage and make it CertMagic's default storage, too
 	err = func() error {

context.go

@@ -393,6 +393,8 @@ func (ctx Context) LoadModuleByID(id string, rawMsg json.RawMessage) (any, error
 		return nil, fmt.Errorf("module value cannot be null")
 	}
 
+	var err error
+
 	// if this is an app module, keep a reference to it,
 	// since submodules may need to reference it during
 	// provisioning (even though the parent app module
@@ -402,12 +404,17 @@ func (ctx Context) LoadModuleByID(id string, rawMsg json.RawMessage) (any, error
 	// module has been configured for DNS challenges)
 	if appModule, ok := val.(App); ok {
 		ctx.cfg.apps[id] = appModule
+		defer func() {
+			if err != nil {
+				ctx.cfg.failedApps[id] = err
+			}
+		}()
 	}
 
 	ctx.ancestry = append(ctx.ancestry, val)
 
 	if prov, ok := val.(Provisioner); ok {
-		err := prov.Provision(ctx)
+		err = prov.Provision(ctx)
 		if err != nil {
 			// incomplete provisioning could have left state
 			// dangling, so make sure it gets cleaned up
@@ -422,7 +429,7 @@ func (ctx Context) LoadModuleByID(id string, rawMsg json.RawMessage) (any, error
 	}
 
 	if validator, ok := val.(Validator); ok {
-		err := validator.Validate()
+		err = validator.Validate()
 		if err != nil {
 			// since the module was already provisioned, make sure we clean up
 			if cleanerUpper, ok := val.(CleanerUpper); ok {
@@ -487,6 +494,10 @@ func (ctx Context) loadModuleInline(moduleNameKey, moduleScope string, raw json.
 // or stop App modules. The caller is expected to assert to the
 // concrete type.
 func (ctx Context) App(name string) (any, error) {
+	// if the app failed to load before, return the cached error
+	if err, ok := ctx.cfg.failedApps[name]; ok {
+		return nil, fmt.Errorf("loading %s app module: %v", name, err)
+	}
 	if app, ok := ctx.cfg.apps[name]; ok {
 		return app, nil
 	}
@@ -511,6 +522,10 @@ func (ctx Context) AppIfConfigured(name string) (any, error) {
 	if ctx.cfg == nil {
 		return nil, fmt.Errorf("app module %s: %w", name, ErrNotConfigured)
 	}
+	// if the app failed to load before, return the cached error
+	if err, ok := ctx.cfg.failedApps[name]; ok {
+		return nil, fmt.Errorf("loading %s app module: %v", name, err)
+	}
 	if app, ok := ctx.cfg.apps[name]; ok {
 		return app, nil
 	}