docker, core, editoast: add mode single-worker for all infra

README.md

@@ -48,7 +48,7 @@ docker compose up -d --build
 xdg-open http://localhost:4000/
 ```
 
-(Linux or WSL users can use `scripts/osrd-compose.sh` instead of `docker compose` to enable host networking - useful to launch services in a debugger)
+(Linux or WSL users can use `scripts/host-compose.sh` instead of `docker compose` to enable host networking - useful to launch services in a debugger)
 
 ## Deployment
 

core/README.md

@@ -40,11 +40,14 @@ You'll need **Java 17**
 # on Windows
 gradlew.bat shadowJar
 
-# Run as service
-java -jar build/libs/osrd-all.jar api --editoast-url http://localhost:8090/ -p 8080
+# Run as a RabbitMQ single worker for all infra
+ALL_INFRA=true java -jar build/libs/osrd-all.jar worker --editoast-url http://localhost:8090/
 
 # Check that an infra can be loaded
 java -jar build/libs/osrd-all.jar load-infra --path RAILJSON_INFRA
+
+# Run as web-service (deprecated inside OSRD's stack)
+java -jar build/libs/osrd-all.jar api --editoast-url http://localhost:8090/ --port 8080
 ```
 
 ### CLI usage (alternative)
@@ -81,3 +84,35 @@ To auto-format all source code, run:
 ```sh
 ./gradlew spotlessApply
 ```
+
+### Local run and debug
+
+It is recommended to pass additional Java options to enable the process of big infra:
+
+```sh
+-ea -Xmx12g -XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=dump.hprof
+```
+
+Using a specific script (just through `docker compose` CLI and a set of docker-compose files)
+allows to run a single core worker for all infra on localhost network:
+
+```sh
+./scripts/single-worker-compose.sh up -d
+
+# or exclude 'core' service straight away:
+./scripts/single-worker-compose.sh up -d --scale core=0
+```
+
+Then, it is easy to replace the desired component for debug purpose. \
+For core:
+
+```sh
+./scripts/single-worker-compose.sh down core # if 'core' is running
+./gradlew shadowJar && ALL_INFRA=true java -jar -ea -Xmx12g -XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=dump.hprof build/libs/osrd-all.jar worker --editoast-url http://localhost:8090/
+```
+
+Clean or restart the whole stack can be necessary sometimes and is also available
+through docker compose CLI (the following wipes the database too):
+```sh
+./scripts/single-worker-compose.sh down -v
+```

core/src/main/java/fr/sncf/osrd/cli/WorkerCommand.kt

@@ -2,7 +2,10 @@ package fr.sncf.osrd.cli
 
 import com.beust.jcommander.Parameter
 import com.beust.jcommander.Parameters
-import com.rabbitmq.client.*
+import com.rabbitmq.client.AMQP
+import com.rabbitmq.client.Channel
+import com.rabbitmq.client.ConnectionFactory
+import com.rabbitmq.client.DeliverCallback
 import fr.sncf.osrd.api.*
 import fr.sncf.osrd.api.api_v2.conflicts.ConflictDetectionEndpointV2
 import fr.sncf.osrd.api.api_v2.path_properties.PathPropEndpoint
@@ -39,16 +42,18 @@ class WorkerCommand : CliCommand {
     private var editoastAuthorization: String = "x-osrd-core"
 
     val WORKER_ID: String?
-    val WORKER_ID_USE_HOSTNAME: String?
+    val WORKER_ID_USE_HOSTNAME: Boolean
     val WORKER_KEY: String?
     val WORKER_AMQP_URI: String
     val WORKER_POOL: String
     val WORKER_REQUESTS_QUEUE: String
     val WORKER_ACTIVITY_EXCHANGE: String
+    val ALL_INFRA: Boolean
 
     init {
-        WORKER_ID_USE_HOSTNAME = System.getenv("WORKER_ID_USE_HOSTNAME")
-        WORKER_KEY = System.getenv("WORKER_KEY")
+        WORKER_ID_USE_HOSTNAME = getBooleanEnvvar("WORKER_ID_USE_HOSTNAME")
+        ALL_INFRA = getBooleanEnvvar("ALL_INFRA")
+        WORKER_KEY = if (ALL_INFRA) "all" else System.getenv("WORKER_KEY")
         WORKER_AMQP_URI =
             System.getenv("WORKER_AMQP_URI") ?: "amqp://osrd:password@127.0.0.1:5672/%2f"
         WORKER_POOL = System.getenv("WORKER_POOL") ?: "core"
@@ -58,18 +63,19 @@ class WorkerCommand : CliCommand {
             System.getenv("WORKER_ACTIVITY_EXCHANGE") ?: "$WORKER_POOL-activity-xchg"
 
         WORKER_ID =
-            if (
-                WORKER_ID_USE_HOSTNAME == null ||
-                    WORKER_ID_USE_HOSTNAME == "" ||
-                    WORKER_ID_USE_HOSTNAME == "0" ||
-                    WORKER_ID_USE_HOSTNAME.lowercase() == "false"
-            ) {
-                System.getenv("WORKER_ID")
-            } else {
+            if (WORKER_ID_USE_HOSTNAME) {
                 java.net.InetAddress.getLocalHost().hostName
+            } else if (ALL_INFRA) {
+                "all_infra_worker"
+            } else {
+                System.getenv("WORKER_ID")
             }
     }
 
+    private fun getBooleanEnvvar(name: String): Boolean {
+        return System.getenv(name)?.lowercase() !in arrayOf(null, "", "0", "false")
+    }
+
     override fun run(): Int {
         if (WORKER_ID == null || WORKER_KEY == null) {
             throw IllegalStateException(
@@ -123,7 +129,9 @@ class WorkerCommand : CliCommand {
         val connection = factory.newConnection()
         connection.createChannel().use { channel -> reportActivity(channel, "started") }
 
-        infraManager.load(infraId, null, diagnosticRecorder)
+        if (!ALL_INFRA) {
+            infraManager.load(infraId, null, diagnosticRecorder)
+        }
 
         connection.createChannel().use { channel -> reportActivity(channel, "ready") }
 

docker-compose.yml

@@ -21,6 +21,7 @@ services:
       test: ["CMD", "pg_isready", "-d", "postgres://osrd:password@postgres/osrd"]
       start_period: 4s
       interval: 5s
+      retries: 20
 
   valkey:
     image: valkey/valkey:alpine
@@ -35,6 +36,7 @@ services:
       test: ["CMD", "valkey-cli", "ping"]
       start_period: 4s
       interval: 5s
+      retries: 20
 
   rabbitmq:
     image: rabbitmq:4-management
@@ -49,20 +51,33 @@ services:
       - rabbitmq_data:/var/lib/rabbitmq
     healthcheck:
       test: ["CMD", "rabbitmqctl", "status"]
-      interval: 30s
-      timeout: 30s
-      retries: 3
+      interval: 5s
+      timeout: 5s
+      retries: 20
 
   core:
     # This is a dummy container to build the core image
+    # and document/provide parameters to other compose files
     image: ghcr.io/openrailassociation/osrd-edge/osrd-core:${TAG-dev}
     container_name: osrd-core-dummy
+    depends_on:
+      rabbitmq: {condition: service_healthy}
     build:
       context: core
       dockerfile: Dockerfile
       additional_contexts:
         test_data: tests/data
         static_assets: assets
+    environment:
+      # Actual values in ./docker/osrdyne.yml (please maintain consistency)
+      # Provided here only for reuse in compose layers and doc
+      CORE_EDITOAST_URL: "http://osrd-editoast"
+      JAVA_TOOL_OPTIONS: "-javaagent:/app/opentelemetry-javaagent.jar"
+      CORE_MONITOR_TYPE: "opentelemetry"
+      OTEL_EXPORTER_OTLP_TRACES_PROTOCOL: "grpc"
+      OTEL_EXPORTER_OTLP_TRACES_ENDPOINT: "http://jaeger:4317"
+      OTEL_METRICS_EXPORTER: "none"
+      OTEL_LOGS_EXPORTER: "none"
     restart: "no"
     command: "true"
 
@@ -88,6 +103,7 @@ services:
       test: ["CMD", "curl", "-f", "http://localhost:3000"]
       start_period: 4s
       interval: 5s
+      retries: 6
 
   editoast:
     image: ghcr.io/openrailassociation/osrd-edge/osrd-editoast:${TAG-dev}
@@ -122,6 +138,7 @@ services:
       test: ["CMD", "curl", "-f", "http://localhost/health"]
       start_period: 4s
       interval: 5s
+      retries: 6
 
   gateway:
     image: ghcr.io/openrailassociation/osrd-edge/osrd-gateway:${TAG-dev}-standalone

docker/docker-compose.host.yml

@@ -16,7 +16,6 @@ services:
     network_mode: host
     environment:
       EDITOAST_PORT: 8090
-      OSRD_BACKEND_URL: "http://localhost:8080"
       OSRDYNE_API_URL: "http://localhost:4242"
       VALKEY_URL: "redis://localhost"
       DATABASE_URL: "postgres://osrd:password@localhost:5432/osrd"
@@ -39,6 +38,13 @@ services:
     ports: []
     network_mode: host
 
+  core:
+    ports: []
+    network_mode: host
+    environment:
+      CORE_EDITOAST_URL: "http://127.0.0.1:8090"
+      OTEL_EXPORTER_OTLP_TRACES_ENDPOINT: "http://127.0.0.1:4317"
+
   osrdyne:
     ports: []
     network_mode: host

docker/docker-compose.pr-tests.yml

@@ -32,6 +32,7 @@ services:
       test: ["CMD", "pg_isready", "-d", "postgres://osrd:password@postgres:5433/osrd"]
       start_period: 4s
       interval: 5s
+      retries: 20
     networks:
       - pr-tests
 
@@ -48,6 +49,7 @@ services:
       test: ["CMD", "valkey-cli", "-p" , "6380", "ping"]
       start_period: 4s
       interval: 5s
+      retries: 20
     networks:
       - pr-tests
 
@@ -67,9 +69,9 @@ services:
       - ../docker/rabbitmq-pr-tests.conf:/etc/rabbitmq/rabbitmq.conf
     healthcheck:
       test: ["CMD", "rabbitmqctl", "status"]
-      interval: 15s
-      timeout: 30s
-      retries: 3
+      interval: 5s
+      timeout: 5s
+      retries: 20
     networks:
       - pr-tests
 
@@ -107,6 +109,7 @@ services:
       test: ["CMD", "curl", "-f", "http://localhost:8091/health"]
       start_period: 4s
       interval: 5s
+      retries: 6
     networks:
       - pr-tests
 

docker/docker-compose.single-worker.yml

@@ -0,0 +1,20 @@
+version: '3'
+
+services:
+  editoast:
+    environment:
+      - EDITOAST_CORE_SINGLE_WORKER=true
+
+  osrdyne:
+    volumes: !reset []
+
+  core:
+    # Launch a single worker handling all infra
+    environment:
+      - ALL_INFRA=true
+    container_name: osrd-core-all-infra
+    restart: unless-stopped
+    command:
+      - /bin/sh
+      - -c
+      - "exec java $JAVA_OPTS -ea -jar /app/osrd_core.jar worker"

editoast/openapi.yaml

@@ -3756,30 +3756,6 @@ components:
           type: string
           enum:
           - editoast:coreclient:BrokenPipe
-    EditoastCoreErrorCannotExtractResponseBody:
-      type: object
-      required:
-      - type
-      - status
-      - message
-      properties:
-        context:
-          type: object
-          required:
-          - msg
-          properties:
-            msg:
-              type: string
-        message:
-          type: string
-        status:
-          type: integer
-          enum:
-          - 500
-        type:
-          type: string
-          enum:
-          - editoast:coreclient:CannotExtractResponseBody
     EditoastCoreErrorConnectionClosedBeforeMessageCompleted:
       type: object
       required:
@@ -4055,7 +4031,6 @@ components:
       - $ref: '#/components/schemas/EditoastCacheOperationErrorDuplicateIdsProvided'
       - $ref: '#/components/schemas/EditoastCacheOperationErrorObjectNotFound'
       - $ref: '#/components/schemas/EditoastCoreErrorBrokenPipe'
-      - $ref: '#/components/schemas/EditoastCoreErrorCannotExtractResponseBody'
       - $ref: '#/components/schemas/EditoastCoreErrorConnectionClosedBeforeMessageCompleted'
       - $ref: '#/components/schemas/EditoastCoreErrorConnectionResetByPeer'
       - $ref: '#/components/schemas/EditoastCoreErrorCoreResponseFormatError'

editoast/src/client/mod.rs

@@ -151,6 +151,8 @@ pub struct RunserverArgs {
     pub mq_url: String,
     #[clap(long, env = "EDITOAST_CORE_TIMEOUT", default_value_t = 180)]
     pub core_timeout: u64,
+    #[clap(long, env = "EDITOAST_CORE_SINGLE_WORKER", default_value_t = false)]
+    pub core_single_worker: bool,
     #[clap(long, env = "ROOT_PATH", default_value_t = String::new())]
     pub root_path: String,
     #[clap(long)]