Feature flagging batch scheduling (#592)

Netflix · Nov 28, 2023 · b577468 · b577468
1 parent b244630
commit b577468
Show file tree

Hide file tree

Showing 2 changed files with 11 additions and 2 deletions.
diff --git a/...mantis-control-plane-server/src/main/java/io/mantisrx/master/jobcluster/job/JobActor.java b/...mantis-control-plane-server/src/main/java/io/mantisrx/master/jobcluster/job/JobActor.java
@@ -1259,6 +1259,7 @@ class WorkerManager implements IWorkerManager {
                 .expireAfterWrite(1, TimeUnit.HOURS)
                 .build();
         private volatile boolean stageAssignmentPotentiallyChanged;
+        private final boolean batchSchedulingEnabled = ConfigurationProvider.getConfig().isBatchSchedulingEnabled();
 
         /**
          * Creates an instance of this class.
@@ -1361,7 +1362,7 @@ private void initializeRunningWorkers() {
 
                         // If the job is in accepted state, queue all its pending workers at once in a batch request.
                         // This is important when before master failover there were pending batch requests
-                        if (JobState.isAcceptedState(mantisJobMetaData.getState())) {
+                        if (batchSchedulingEnabled && JobState.isAcceptedState(mantisJobMetaData.getState())) {
                             workersToSubmit.add(wm);
                         } else {
                             queueTask(wm);
@@ -1495,7 +1496,11 @@ private void submitInitialWorkers() throws Exception {
 
                 if (!workers.isEmpty()) {
                     // queue to scheduler
-                    queueTasks(workers, empty());
+                    if (batchSchedulingEnabled) {
+                        workers.forEach(this::queueTask);
+                    } else {
+                        queueTasks(workers, empty());
+                    }
                 }
             } catch (Exception e) {
                 LOGGER.error("Error {} storing workers of job {}", e.getMessage(), jobId.getId(), e);

diff --git a/...trol-plane-server/src/main/java/io/mantisrx/server/master/config/MasterConfiguration.java b/...trol-plane-server/src/main/java/io/mantisrx/server/master/config/MasterConfiguration.java
@@ -387,6 +387,10 @@ default Duration getSchedulerIntervalBetweenRetries() {
     @Default("5000")
     int getResourceClusterActionsPermitsPerSecond();
 
+    @Config("mantis.batchScheduling.enabled")
+    @Default("false")
+    boolean isBatchSchedulingEnabled();
+
     default Duration getHeartbeatInterval() {
         return Duration.ofMillis(getHeartbeatIntervalInMs());
     }