refactor(amber): rename execution coordinators to managers

Yicong-Huang · claude · Yicong-Huang · commit b6abd55c9911 · 2026-07-04T17:52:30.000-07:00
Co-Authored-By: Claude Fable 5 &lt;noreply@anthropic.com&gt;
diff --git a/amber/src/main/python/core/architecture/packaging/output_manager.py b/amber/src/main/python/core/architecture/packaging/output_manager.py
@@ -97,7 +97,7 @@ def is_missing_output_ports(self):
         This method is only used for ensuring correct region execution.
         Some operators may have input port dependency relationships, for
         which we currently use a two-phase region execution scheme.
-        (See `RegionExecutionCoordinator.scala` for details.)
+        (See `RegionExecutionManager.scala` for details.)
         This logic will only be executed when the worker is part of an
         `executingDependeePortPhase` region-execution phase.
         We currently assume that in this phase the operator (worker) will
diff --git a/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/Controller.scala b/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/Controller.scala
@@ -116,7 +116,7 @@ class Controller(
   override def initState(): Unit = {
     attachRuntimeServicesToCPState()
     cp.workflowScheduler.updateSchedule(physicalPlan)
-    cp.workflowExecutionCoordinator.schedule = cp.workflowScheduler.getSchedule
+    cp.workflowExecutionManager.schedule = cp.workflowScheduler.getSchedule
 
     val regions: List[(Long, List[String])] =
       cp.workflowScheduler.getSchedule.getRegions.map { region =>
diff --git a/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/ControllerProcessor.scala b/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/ControllerProcessor.scala
@@ -29,7 +29,7 @@ import org.apache.texera.amber.engine.architecture.common.{
 }
 import org.apache.texera.amber.engine.architecture.controller.execution.WorkflowExecution
 import org.apache.texera.amber.engine.architecture.logreplay.ReplayLogManager
-import org.apache.texera.amber.engine.architecture.scheduling.WorkflowExecutionCoordinator
+import org.apache.texera.amber.engine.architecture.scheduling.WorkflowExecutionManager
 import org.apache.texera.amber.engine.architecture.worker.WorkflowWorker.MainThreadDelegateMessage
 import org.apache.texera.amber.engine.common.ambermessage.WorkflowFIFOMessage
 
@@ -43,7 +43,7 @@ class ControllerProcessor(
   val workflowExecution: WorkflowExecution = WorkflowExecution()
   val workflowScheduler: WorkflowScheduler =
     new WorkflowScheduler(workflowContext, actorId)
-  val workflowExecutionCoordinator: WorkflowExecutionCoordinator = new WorkflowExecutionCoordinator(
+  val workflowExecutionManager: WorkflowExecutionManager = new WorkflowExecutionManager(
     workflowExecution,
     controllerConfig,
     asyncRPCClient
@@ -73,7 +73,7 @@ class ControllerProcessor(
 
   def setupActorRefService(actorRefService: PekkoActorRefMappingService): Unit = {
     this.actorRefService = actorRefService
-    this.workflowExecutionCoordinator.setupActorRefService(this.actorRefService)
+    this.workflowExecutionManager.setupActorRefService(this.actorRefService)
   }
 
   @transient var logManager: ReplayLogManager = _
diff --git a/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/promisehandlers/JumpToOperatorRegionHandler.scala b/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/promisehandlers/JumpToOperatorRegionHandler.scala
@@ -34,15 +34,15 @@ trait JumpToOperatorRegionHandler {
       msg: JumpToOperatorRegionRequest,
       ctx: AsyncRPCContext
   ): Future[EmptyReturn] = {
-    val coordinator = cp.workflowExecutionCoordinator
-    coordinator.schedule.levelSets
+    val manager = cp.workflowExecutionManager
+    manager.schedule.levelSets
       .collectFirst {
         case (level, regions)
             if regions.exists(_.getOperators.exists(_.id.logicalOpId == msg.targetOperatorId)) =>
           level
       }
       .foreach { targetLevel =>
-        coordinator.schedule = coordinator.schedule.copy(initialLevelIndex = targetLevel)
+        manager.schedule = manager.schedule.copy(initialLevelIndex = targetLevel)
       }
     EmptyReturn()
   }
diff --git a/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/promisehandlers/LinkWorkersHandler.scala b/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/promisehandlers/LinkWorkersHandler.scala
@@ -38,7 +38,7 @@ trait LinkWorkersHandler {
   this: ControllerAsyncRPCHandlerInitializer =>
 
   override def linkWorkers(msg: LinkWorkersRequest, ctx: AsyncRPCContext): Future[EmptyReturn] = {
-    val region = cp.workflowExecutionCoordinator.getRegionOfLink(msg.link)
+    val region = cp.workflowExecutionManager.getRegionOfLink(msg.link)
     val resourceConfig = region.resourceConfig.get
     val linkConfig = resourceConfig.linkConfigs(msg.link)
     val linkExecution =
diff --git a/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/promisehandlers/PortCompletedHandler.scala b/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/promisehandlers/PortCompletedHandler.scala
@@ -64,7 +64,7 @@ trait PortCompletedHandler {
           msg.portId,
           input = msg.input
         )
-        cp.workflowExecutionCoordinator.getRegionOfPortId(globalPortId) match {
+        cp.workflowExecutionManager.getRegionOfPortId(globalPortId) match {
           case Some(region) =>
             val regionExecution = cp.workflowExecution.getRegionExecution(region.id)
             val operatorExecution =
@@ -81,8 +81,8 @@ trait PortCompletedHandler {
               else operatorExecution.isOutputPortCompleted(msg.portId)
 
             if (isPortCompleted) {
-              cp.workflowExecutionCoordinator
-                .coordinateRegionExecutors(cp.actorService)
+              cp.workflowExecutionManager
+                .advanceRegionExecutions(cp.actorService)
                 // Since this message is sent from a worker, any exception from the above code will be returned to that worker.
                 // Additionally, a fatal error is sent to the client, indicating that the region cannot be scheduled.
                 .onFailure {
diff --git a/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/promisehandlers/ReconfigurationHandler.scala b/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/promisehandlers/ReconfigurationHandler.scala
@@ -61,7 +61,7 @@ trait ReconfigurationHandler {
     }
     val futures = mutable.ArrayBuffer[Future[_]]()
     val friesComponents =
-      FriesReconfigurationAlgorithm.getReconfigurations(cp.workflowExecutionCoordinator, msg)
+      FriesReconfigurationAlgorithm.getReconfigurations(cp.workflowExecutionManager, msg)
     friesComponents.foreach { friesComponent =>
       if (friesComponent.scope.size == 1) {
         val updateExecutorRequest = friesComponent.reconfigurations.head
diff --git a/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/promisehandlers/StartWorkflowHandler.scala b/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/promisehandlers/StartWorkflowHandler.scala
@@ -41,8 +41,8 @@ trait StartWorkflowHandler {
       ctx: AsyncRPCContext
   ): Future[StartWorkflowResponse] = {
     if (cp.workflowExecution.getState.isUninitialized) {
-      cp.workflowExecutionCoordinator
-        .coordinateRegionExecutors(cp.actorService)
+      cp.workflowExecutionManager
+        .advanceRegionExecutions(cp.actorService)
         .map(_ => {
           cp.controllerTimerService.enableStatusUpdate()
           cp.controllerTimerService.enableRuntimeStatisticsCollection()
diff --git a/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/promisehandlers/WorkerExecutionCompletedHandler.scala b/amber/src/main/scala/org/apache/texera/amber/engine/architecture/controller/promisehandlers/WorkerExecutionCompletedHandler.scala
@@ -64,7 +64,7 @@ trait WorkerExecutionCompletedHandler {
         val isWorkflowTerminal =
           cp.workflowExecution.isCompleted &&
             !cp.workflowScheduler.hasPendingRegions &&
-            !cp.workflowExecutionCoordinator.hasUnfinishedRegionCoordinators
+            !cp.workflowExecutionManager.hasUnfinishedRegionManagers
         if (isWorkflowTerminal) {
           // after query result come back: send completed event, cleanup ,and kill workflow
           sendToClient(ExecutionStateUpdate(cp.workflowExecution.getState))
diff --git a/amber/src/main/scala/org/apache/texera/amber/engine/architecture/messaginglayer/OutputManager.scala b/amber/src/main/scala/org/apache/texera/amber/engine/architecture/messaginglayer/OutputManager.scala
@@ -285,7 +285,7 @@ class OutputManager(
 
   /**
     * This method is only used for ensuring correct region execution. Some operators may have input port dependency
-    * relationships, for which we currently use a two-phase region execution scheme.  (See `RegionExecutionCoordinator`
+    * relationships, for which we currently use a two-phase region execution scheme.  (See `RegionExecutionManager`
     * for details.)
     * This logic will only be executed when the worker is part of an `executingDependeePort` region-execution phase.
     * We currently assume that in this phase the operator (worker) will not output any data, hence no output ports.
@@ -322,7 +322,7 @@ class OutputManager(
     writerThread.start()
 
     // The state document is provisioned alongside the result document
-    // by RegionExecutionCoordinator, so it is always present.
+    // by RegionExecutionManager, so it is always present.
     val stateWriter = DocumentFactory
       .openDocument(VFSURIFactory.stateURI(portBaseURI))
       ._1
diff --git a/amber/src/main/scala/org/apache/texera/amber/engine/architecture/scheduling/RegionExecutionManager.scala b/amber/src/main/scala/org/apache/texera/amber/engine/architecture/scheduling/RegionExecutionManager.scala
@@ -64,7 +64,7 @@ import java.util.concurrent.TimeUnit
 import java.util.concurrent.atomic.AtomicReference
 import scala.concurrent.duration.{Duration => ScalaDuration}
 
-object RegionExecutionCoordinator {
+object RegionExecutionManager {
 
   // Max EndWorker retries before termination fails. ~30s at DefaultKillRetryDelay (200ms).
   private[scheduling] val DefaultMaxTerminationAttempts: Int = 150
@@ -99,16 +99,16 @@ object RegionExecutionCoordinator {
   *
   * 3. `Completed`
   */
-class RegionExecutionCoordinator(
+class RegionExecutionManager(
     region: Region,
     isRestart: Boolean,
     workflowExecution: WorkflowExecution,
     asyncRPCClient: AsyncRPCClient,
     controllerConfig: ControllerConfig,
     actorService: PekkoActorService,
     actorRefService: PekkoActorRefMappingService,
-    maxTerminationAttempts: Int = RegionExecutionCoordinator.DefaultMaxTerminationAttempts,
-    killRetryDelay: TwitterDuration = RegionExecutionCoordinator.DefaultKillRetryDelay
+    maxTerminationAttempts: Int = RegionExecutionManager.DefaultMaxTerminationAttempts,
+    killRetryDelay: TwitterDuration = RegionExecutionManager.DefaultKillRetryDelay
 ) extends AmberLogging {
 
   initRegionExecution()
@@ -126,8 +126,8 @@ class RegionExecutionCoordinator(
   private val killRetryTimer: Timer = new JavaTimer(true)
 
   /**
-    * Sync the status of `RegionExecution` and transition this coordinator's phase to `Completed` only when the
-    * coordinator is currently in `ExecutingNonDependeePortsPhase`, all the ports of this region are completed, and
+    * Sync the status of `RegionExecution` and transition this manager's phase to `Completed` only when the
+    * manager is currently in `ExecutingNonDependeePortsPhase`, all the ports of this region are completed, and
     * all workers in this region are terminated.
     *
     * Additionally, this method will also terminate all the workers of this region:
@@ -156,8 +156,8 @@ class RegionExecutionCoordinator(
       existingTerminationFuture
     } else {
       val terminationFuture = terminateWorkersWithRetry(regionExecution).flatMap { _ =>
-        // Set this coordinator's status to be completed so that subsequent regions can be started by
-        // WorkflowExecutionCoordinator.
+        // Set this manager's status to be completed so that subsequent regions can be started by
+        // WorkflowExecutionManager.
         setPhase(Completed)
         Future.Unit
       }
diff --git a/amber/src/main/scala/org/apache/texera/amber/engine/architecture/scheduling/WorkflowExecutionManager.scala b/amber/src/main/scala/org/apache/texera/amber/engine/architecture/scheduling/WorkflowExecutionManager.scala
@@ -34,7 +34,7 @@ import org.apache.texera.amber.engine.common.rpc.AsyncRPCClient
 import java.util.concurrent.atomic.AtomicBoolean
 import scala.collection.mutable
 
-class WorkflowExecutionCoordinator(
+class WorkflowExecutionManager(
     workflowExecution: WorkflowExecution,
     controllerConfig: ControllerConfig,
     asyncRPCClient: AsyncRPCClient
@@ -44,8 +44,7 @@ class WorkflowExecutionCoordinator(
 
   private val executedRegions: mutable.ListBuffer[Set[Region]] = mutable.ListBuffer()
 
-  private val regionExecutionCoordinators
-      : mutable.HashMap[RegionIdentity, RegionExecutionCoordinator] =
+  private val regionExecutionManagers: mutable.HashMap[RegionIdentity, RegionExecutionManager] =
     mutable.HashMap()
   private val completionNotified: AtomicBoolean = new AtomicBoolean(false)
 
@@ -56,29 +55,29 @@ class WorkflowExecutionCoordinator(
   }
 
   /**
-    * Each invocation first syncs the internal statuses of each exisiting `RegionExecutionCoordintor`, after which each
-    * of the `RegionExecutionCoordintor`s will launch the corresponding next phase of whenever needed until it is
+    * Each invocation first syncs the internal statuses of each exisiting `RegionExecutionManager`, after which each
+    * of the `RegionExecutionManager`s will launch the corresponding next phase of whenever needed until it is
     * in `Completed` status (phase).
     *
     * After the syncs, if there are no running region(s), it will start new regions (if available).
     */
-  def coordinateRegionExecutors(actorService: PekkoActorService): Future[Unit] = {
-    val unfinishedRegionCoordinators =
-      regionExecutionCoordinators.values.filter(!_.isCompleted).toSeq
+  def advanceRegionExecutions(actorService: PekkoActorService): Future[Unit] = {
+    val unfinishedRegionManagers =
+      regionExecutionManagers.values.filter(!_.isCompleted).toSeq
 
     // Trigger sync for each unfinished region.
-    unfinishedRegionCoordinators.foreach(_.syncStatusAndTransitionRegionExecutionPhase())
+    unfinishedRegionManagers.foreach(_.syncStatusAndTransitionRegionExecutionPhase())
 
-    // Wait only for region termination futures (kill path), then re-run coordination.
-    val terminationFutures = unfinishedRegionCoordinators.flatMap(_.getTerminationFutureOpt)
+    // Wait only for region termination futures (kill path), then re-run the advance loop.
+    val terminationFutures = unfinishedRegionManagers.flatMap(_.getTerminationFutureOpt)
     if (terminationFutures.nonEmpty) {
       return Future
         .collect(terminationFutures)
         .unit
-        .flatMap(_ => coordinateRegionExecutors(actorService))
+        .flatMap(_ => advanceRegionExecutions(actorService))
     }
 
-    if (regionExecutionCoordinators.values.exists(!_.isCompleted)) {
+    if (regionExecutionManagers.values.exists(!_.isCompleted)) {
       // Some regions are still not completed yet. Cannot start the new regions.
       return Future.Unit
     }
@@ -103,7 +102,7 @@ class WorkflowExecutionCoordinator(
             } else {
               workflowExecution.initRegionExecution(region)
             }
-            regionExecutionCoordinators(region.id) = new RegionExecutionCoordinator(
+            regionExecutionManagers(region.id) = new RegionExecutionManager(
               region,
               isRestart,
               workflowExecution,
@@ -112,7 +111,7 @@ class WorkflowExecutionCoordinator(
               actorService,
               actorRefService
             )
-            regionExecutionCoordinators(region.id)
+            regionExecutionManagers(region.id)
           })
           .map(_.syncStatusAndTransitionRegionExecutionPhase())
           .toSeq
@@ -134,8 +133,8 @@ class WorkflowExecutionCoordinator(
       .toSet
   }
 
-  def hasUnfinishedRegionCoordinators: Boolean = {
-    regionExecutionCoordinators.values.exists(!_.isCompleted)
+  def hasUnfinishedRegionManagers: Boolean = {
+    regionExecutionManagers.values.exists(!_.isCompleted)
   }
 
 }
diff --git a/amber/src/main/scala/org/apache/texera/amber/engine/common/FriesReconfigurationAlgorithm.scala b/amber/src/main/scala/org/apache/texera/amber/engine/common/FriesReconfigurationAlgorithm.scala
@@ -25,7 +25,7 @@ import org.apache.texera.amber.engine.architecture.rpc.controlcommands.{
   UpdateExecutorRequest,
   WorkflowReconfigureRequest
 }
-import org.apache.texera.amber.engine.architecture.scheduling.{Region, WorkflowExecutionCoordinator}
+import org.apache.texera.amber.engine.architecture.scheduling.{Region, WorkflowExecutionManager}
 import org.jgrapht.alg.connectivity.ConnectivityInspector
 
 import scala.collection.mutable
@@ -45,11 +45,11 @@ object FriesReconfigurationAlgorithm {
   }
 
   def getReconfigurations(
-      workflowExecutionCoordinator: WorkflowExecutionCoordinator,
+      workflowExecutionManager: WorkflowExecutionManager,
       reconfiguration: WorkflowReconfigureRequest
   ): Set[FriesComponent] = {
     // independently schedule reconfigurations for each region:
-    workflowExecutionCoordinator.getExecutingRegions
+    workflowExecutionManager.getExecutingRegions
       .flatMap(region => computeMCS(region, reconfiguration, reconfiguration.reconfigurationId))
   }
 
diff --git a/amber/src/main/scala/org/apache/texera/web/resource/SyncExecutionResource.scala b/amber/src/main/scala/org/apache/texera/web/resource/SyncExecutionResource.scala
@@ -281,7 +281,7 @@ class SyncExecutionResource extends LazyLogging {
             killExecution(executionService)
             (executionService.executionStateStore.metadataStore.getState, true, false)
           case TargetResultsReady(_) =>
-            // RegionExecutionCoordinator caches upstream results asynchronously after operators
+            // RegionExecutionManager caches upstream results asynchronously after operators
             // complete; sleep gives that caching a chance to finish before we shut down the client.
             // TODO: replace with a synchronous signal from the engine.
             Thread.sleep(500)
diff --git a/amber/src/test/python/core/architecture/packaging/test_state_materialization_e2e.py b/amber/src/test/python/core/architecture/packaging/test_state_materialization_e2e.py
@@ -162,7 +162,7 @@ def test_state_written_by_output_manager_is_replayed_by_reader():
     port_id = PortIdentity(id=0, internal=False)
     worker_schema_for_result = State.SCHEMA  # producer-side: only state matters
 
-    # 1. RegionExecutionCoordinator's responsibility: provision result +
+    # 1. RegionExecutionManager's responsibility: provision result +
     # state documents at the port base URI before any worker starts.
     # We emulate that here.
     DocumentFactory.create_document(
diff --git a/amber/src/test/scala/org/apache/texera/amber/engine/architecture/scheduling/RegionExecutionManagerSpec.scala b/amber/src/test/scala/org/apache/texera/amber/engine/architecture/scheduling/RegionExecutionManagerSpec.scala
diff --git a/amber/src/test/scala/org/apache/texera/amber/engine/architecture/scheduling/RegionExecutionManagerTestSupport.scala b/amber/src/test/scala/org/apache/texera/amber/engine/architecture/scheduling/RegionExecutionManagerTestSupport.scala
diff --git a/amber/src/test/scala/org/apache/texera/amber/engine/architecture/scheduling/WorkflowExecutionManagerSpec.scala b/amber/src/test/scala/org/apache/texera/amber/engine/architecture/scheduling/WorkflowExecutionManagerSpec.scala
diff --git a/amber/src/test/scala/org/apache/texera/amber/engine/architecture/worker/promisehandlers/EndHandlerSpec.scala b/amber/src/test/scala/org/apache/texera/amber/engine/architecture/worker/promisehandlers/EndHandlerSpec.scala

Original file line number	Diff line number	Diff line change
`@@ -34,15 +34,15 @@ trait JumpToOperatorRegionHandler {`
`34`	`34`	`msg: JumpToOperatorRegionRequest,`
`35`	`35`	`ctx: AsyncRPCContext`
`36`	`36`	`): Future[EmptyReturn] = {`
`37`		`- val coordinator = cp.workflowExecutionCoordinator`
`38`		`- coordinator.schedule.levelSets`
	`37`	`+ val manager = cp.workflowExecutionManager`
	`38`	`+ manager.schedule.levelSets`
`39`	`39`	`.collectFirst {`
`40`	`40`	`case (level, regions)`
`41`	`41`	`if regions.exists(_.getOperators.exists(_.id.logicalOpId == msg.targetOperatorId)) =>`
`42`	`42`	`level`
`43`	`43`	`}`
`44`	`44`	`.foreach { targetLevel =>`
`45`		`- coordinator.schedule = coordinator.schedule.copy(initialLevelIndex = targetLevel)`
	`45`	`+ manager.schedule = manager.schedule.copy(initialLevelIndex = targetLevel)`
`46`	`46`	`}`
`47`	`47`	`EmptyReturn()`
`48`	`48`	`}`
Original file line number	Diff line number	Diff line change
`@@ -61,7 +61,7 @@ trait ReconfigurationHandler {`
`61`	`61`	`}`
`62`	`62`	`val futures = mutable.ArrayBuffer[Future[_]]()`
`63`	`63`	`val friesComponents =`
`64`		`- FriesReconfigurationAlgorithm.getReconfigurations(cp.workflowExecutionCoordinator, msg)`
	`64`	`+ FriesReconfigurationAlgorithm.getReconfigurations(cp.workflowExecutionManager, msg)`
`65`	`65`	`friesComponents.foreach { friesComponent =>`
`66`	`66`	`if (friesComponent.scope.size == 1) {`
`67`	`67`	`val updateExecutorRequest = friesComponent.reconfigurations.head`