configure worker pooling via appfile

cli/daemon/run/runtime_config2.go

@@ -49,6 +49,7 @@ type RuntimeConfigGenerator struct {
 		PlatformID() string
 		PlatformOrLocalID() string
 		GlobalCORS() (appfile.CORS, error)
+		BuildSettings() (appfile.Build, error)
 	}
 
 	// The infra manager to use
@@ -82,6 +83,7 @@ type RuntimeConfigGenerator struct {
 	SvcConfigs map[string]string
 
 	conf     *rtconfgen.Builder
+	compute  *runtimev1.Compute
 	authKeys []*runtimev1.EncoreAuthKey
 }
 
@@ -139,6 +141,17 @@ func (g *RuntimeConfigGenerator) initialize() error {
 			})
 		}
 
+		buildSettings, err := g.app.BuildSettings()
+		if err != nil {
+			return errors.Wrap(err, "failed to get app's build settings")
+		}
+		g.compute = &runtimev1.Compute{}
+		if buildSettings.WorkerPooling {
+			// Auto-detect the number of worker threads.
+			n := int32(0)
+			g.compute.WorkerThreads = &n
+		}
+
 		g.conf.AuthMethods([]*runtimev1.ServiceAuth{
 			{
 				AuthMethod: &runtimev1.ServiceAuth_EncoreAuth_{
@@ -436,6 +449,7 @@ func (g *RuntimeConfigGenerator) ProcPerService(proxy *svcproxy.SvcProxy) (servi
 	// Set up the service processes.
 	for _, svc := range g.md.Svcs {
 		conf, err := g.conf.Deployment(newRid()).
+			Compute(g.compute).
 			ServiceDiscovery(sd).
 			HostsServices(svc.Name).
 			ReduceWithMeta(g.md).
@@ -459,7 +473,7 @@ func (g *RuntimeConfigGenerator) ProcPerService(proxy *svcproxy.SvcProxy) (servi
 
 	// Set up the gateways.
 	for _, gw := range g.md.Gateways {
-		conf, err := g.conf.Deployment(newRid()).ServiceDiscovery(sd).HostsGateways(gw.EncoreName).ReduceWithMeta(g.md).BuildRuntimeConfig()
+		conf, err := g.conf.Deployment(newRid()).Compute(g.compute).ServiceDiscovery(sd).HostsGateways(gw.EncoreName).ReduceWithMeta(g.md).BuildRuntimeConfig()
 		if err != nil {
 			return nil, nil, errors.Wrap(err, "failed to generate runtime config")
 		}
@@ -486,7 +500,7 @@ func (g *RuntimeConfigGenerator) AllInOneProc() (*ProcConfig, error) {
 
 	sd := &runtimev1.ServiceDiscovery{Services: make(map[string]*runtimev1.ServiceDiscovery_Location)}
 
-	d := g.conf.Deployment(newRid()).ServiceDiscovery(sd)
+	d := g.conf.Deployment(newRid()).Compute(g.compute).ServiceDiscovery(sd)
 	for _, gw := range g.md.Gateways {
 		d.HostsGateways(gw.EncoreName)
 	}
@@ -556,6 +570,7 @@ func (g *RuntimeConfigGenerator) ProcPerServiceWithNewRuntimeConfig(proxy *svcpr
 
 	for _, svc := range g.md.Svcs {
 		conf, err = g.conf.Deployment(newRid()).
+			Compute(g.compute).
 			ServiceDiscovery(sd).
 			HostsServices(svc.Name).
 			ReduceWithMeta(g.md).
@@ -579,6 +594,7 @@ func (g *RuntimeConfigGenerator) ProcPerServiceWithNewRuntimeConfig(proxy *svcpr
 		}
 
 		conf, err = g.conf.Deployment(newRid()).
+			Compute(g.compute).
 			ServiceDiscovery(sd).
 			HostsGateways(gw.EncoreName).
 			//ReduceWithMeta(g.md).
@@ -604,7 +620,7 @@ func (g *RuntimeConfigGenerator) ForTests(newRuntimeConf bool) (envs []string, e
 
 	sd := &runtimev1.ServiceDiscovery{Services: make(map[string]*runtimev1.ServiceDiscovery_Location)}
 
-	d := g.conf.Deployment(newRid()).ServiceDiscovery(sd)
+	d := g.conf.Deployment(newRid()).Compute(g.compute).ServiceDiscovery(sd)
 	for _, gw := range g.md.Gateways {
 		d.HostsGateways(gw.EncoreName)
 	}

pkg/appfile/appfile.go

@@ -69,6 +69,9 @@ type Build struct {
 	// Docker configures the docker images built
 	// by Encore's CI/CD system.
 	Docker Docker `json:"docker,omitempty"`
+
+	// WorkerPooling enables worker pooling for Encore.ts.
+	WorkerPooling bool `json:"worker_pooling,omitempty"`
 }
 
 type Docker struct {

pkg/rtconfgen/base_builder.go

@@ -135,6 +135,9 @@ type Deployment struct {
 	// The service-discovery configuration for this deployment.
 	sd *runtimev1.ServiceDiscovery
 
+	// The compute configuration for this deployment.
+	compute *runtimev1.Compute
+
 	// The base URL for reaching this deployment from another service.
 	svc2svcBaseURL string
 
@@ -185,6 +188,11 @@ func (d *Deployment) ServiceDiscovery(sd *runtimev1.ServiceDiscovery) *Deploymen
 	return d
 }
 
+func (d *Deployment) Compute(c *runtimev1.Compute) *Deployment {
+	d.compute = c
+	return d
+}
+
 func (d *Deployment) ReduceWithMeta(md *meta.Data) *Deployment {
 	d.reduceWith = option.Some(md)
 	return d
@@ -228,6 +236,7 @@ func (d *Deployment) BuildRuntimeConfig() (*runtimev1.RuntimeConfig, error) {
 		HostedGateways:     gatewayRids,
 		HostedServices:     hostedServices,
 		ServiceDiscovery:   d.sd,
+		Compute:            d.compute,
 		GracefulShutdown:   graceful,
 		DynamicExperiments: d.dynamicExperiments,
 		Observability:      b.obs,

proto/encore/runtime/v1/runtime.proto

@@ -76,23 +76,24 @@ message Deployment {
 }
 
 message Compute {
-   // Number of worker threads to use.
-   // If unset, the runtime automatically determines the number of threads to use
-   // based on the number of CPUs available.
-   optional int32 worker_threads = 1;
-
-   // The minimum log level to use.
-   // If unset it defaults to LOG_LEVEL_TRACE.
-   optional LogLevel log_level = 2;
-
-   enum LogLevel {
-     LOG_LEVEL_DISABLED = 0;
-     LOG_LEVEL_ERROR = 1;
-     LOG_LEVEL_WARN = 2;
-     LOG_LEVEL_INFO = 3;
-     LOG_LEVEL_DEBUG = 4;
-     LOG_LEVEL_TRACE = 5;
-   }
+  // Number of worker threads to use.
+  // If unset it defaults to 1. If set to 0 the runtime
+  // automatically determines the number of threads to use
+  // based on the number of CPUs available.
+  optional int32 worker_threads = 1;
+
+  // The minimum log level to use.
+  // If unset it defaults to LOG_LEVEL_TRACE.
+  optional LogLevel log_level = 2;
+
+  enum LogLevel {
+    LOG_LEVEL_DISABLED = 0;
+    LOG_LEVEL_ERROR = 1;
+    LOG_LEVEL_WARN = 2;
+    LOG_LEVEL_INFO = 3;
+    LOG_LEVEL_DEBUG = 4;
+    LOG_LEVEL_TRACE = 5;
+  }
 }
 
 message Observability {

runtimes/js/encore.dev/internal/appinit/mod.ts

@@ -28,10 +28,10 @@ export function registerGateways(gateways: Gateway[]) {
 
 export async function run(entrypoint: string) {
   if (isMainThread) {
-    const numWorkers = runtime.RT.numExtraWorkerThreads();
-    if (numWorkers > 0) {
+    const extraWorkers = runtime.RT.numWorkerThreads() - 1;
+    if (extraWorkers > 0) {
       const path = fileURLToPath(entrypoint);
-      for (let i = 0; i < numWorkers; i++) {
+      for (let i = 0; i < extraWorkers; i++) {
         new Worker(path);
       }
     }

runtimes/js/src/runtime.rs

@@ -305,12 +305,20 @@ impl Runtime {
         md.clone().into()
     }
 
+    /// Reports the total number of worker threads,
+    /// including the main thread.
     #[napi]
-    pub fn num_extra_worker_threads(&self) -> usize {
-        self.runtime
-            .compute()
-            .worker_threads
-            .map_or_else(|| num_cpus::get() - 1, |v| v as usize)
+    pub fn num_worker_threads(&self) -> u32 {
+        match self.runtime.compute().worker_threads {
+            Some(n) => {
+                if n > 0 {
+                    n as u32
+                } else {
+                    num_cpus::get() as u32
+                }
+            }
+            None => 1u32,
+        }
     }
 }