Avoid hoarding cluster state references during rollover (#124107) (#124257)

By keeping a list of all the rollover results in a rollover request
batch, we were keeping references to all the intermediate cluster states
that we built. We've seen this list take up ~1.4GB with 600 rollover
requests in one batch.

We only kept the list of results to compute the "reason" for the
allocation reroute, so we can easily drop the cluster state reference
from the list and only keep what we need.

Fixes #123893

Author: nielsbauman

docs/changelog/124107.yaml

@@ -0,0 +1,6 @@
+pr: 124107
+summary: Avoid hoarding cluster state references during rollover
+area: Indices APIs
+type: bug
+issues:
+ - 123893

server/src/main/java/org/elasticsearch/action/admin/indices/rollover/LazyRolloverAction.java

@@ -32,7 +32,6 @@
 import org.elasticsearch.cluster.service.MasterServiceTaskQueue;
 import org.elasticsearch.common.Priority;
 import org.elasticsearch.common.Strings;
-import org.elasticsearch.common.collect.Iterators;
 import org.elasticsearch.injection.guice.Inject;
 import org.elasticsearch.tasks.CancellableTask;
 import org.elasticsearch.tasks.Task;
@@ -183,7 +182,7 @@ record LazyRolloverExecutor(
         @Override
         public ClusterState execute(BatchExecutionContext<LazyRolloverTask> batchExecutionContext) {
             final var listener = new AllocationActionMultiListener<RolloverResponse>(threadPool.getThreadContext());
-            final var results = new ArrayList<MetadataRolloverService.RolloverResult>(batchExecutionContext.taskContexts().size());
+            final var results = new ArrayList<String>(batchExecutionContext.taskContexts().size());
             var state = batchExecutionContext.initialState();
             Map<RolloverRequest, List<TaskContext<LazyRolloverTask>>> groupedRequests = new HashMap<>();
             for (final var taskContext : batchExecutionContext.taskContexts()) {
@@ -203,14 +202,7 @@ public ClusterState execute(BatchExecutionContext<LazyRolloverTask> batchExecuti
 
             if (state != batchExecutionContext.initialState()) {
                 var reason = new StringBuilder();
-                Strings.collectionToDelimitedStringWithLimit(
-                    (Iterable<String>) () -> Iterators.map(results.iterator(), t -> t.sourceIndexName() + "->" + t.rolloverIndexName()),
-                    ",",
-                    "lazy bulk rollover [",
-                    "]",
-                    1024,
-                    reason
-                );
+                Strings.collectionToDelimitedStringWithLimit(results, ",", "lazy bulk rollover [", "]", 1024, reason);
                 try (var ignored = batchExecutionContext.dropHeadersContext()) {
                     state = allocationService.reroute(state, reason.toString(), listener.reroute());
                 }
@@ -223,7 +215,7 @@ public ClusterState execute(BatchExecutionContext<LazyRolloverTask> batchExecuti
         public ClusterState executeTask(
             ClusterState currentState,
             RolloverRequest rolloverRequest,
-            List<MetadataRolloverService.RolloverResult> results,
+            ArrayList<String> results,
             List<TaskContext<LazyRolloverTask>> rolloverTaskContexts,
             AllocationActionMultiListener<RolloverResponse> allocationActionMultiListener
         ) throws Exception {
@@ -260,7 +252,7 @@ public ClusterState executeTask(
                 null,
                 isFailureStoreRollover
             );
-            results.add(rolloverResult);
+            results.add(rolloverResult.sourceIndexName() + "->" + rolloverResult.rolloverIndexName());
             logger.trace("lazy rollover result [{}]", rolloverResult);
 
             final var rolloverIndexName = rolloverResult.rolloverIndexName();

server/src/main/java/org/elasticsearch/action/admin/indices/rollover/TransportRolloverAction.java

@@ -47,7 +47,6 @@
 import org.elasticsearch.cluster.service.MasterServiceTaskQueue;
 import org.elasticsearch.common.Priority;
 import org.elasticsearch.common.Strings;
-import org.elasticsearch.common.collect.Iterators;
 import org.elasticsearch.common.unit.ByteSizeValue;
 import org.elasticsearch.common.util.concurrent.EsExecutors;
 import org.elasticsearch.core.Nullable;
@@ -483,7 +482,7 @@ record RolloverExecutor(
         @Override
         public ClusterState execute(BatchExecutionContext<RolloverTask> batchExecutionContext) {
             final var listener = new AllocationActionMultiListener<RolloverResponse>(threadPool.getThreadContext());
-            final var results = new ArrayList<MetadataRolloverService.RolloverResult>(batchExecutionContext.taskContexts().size());
+            final var results = new ArrayList<String>(batchExecutionContext.taskContexts().size());
             var state = batchExecutionContext.initialState();
             for (final var taskContext : batchExecutionContext.taskContexts()) {
                 try (var ignored = taskContext.captureResponseHeaders()) {
@@ -495,14 +494,7 @@ public ClusterState execute(BatchExecutionContext<RolloverTask> batchExecutionCo
 
             if (state != batchExecutionContext.initialState()) {
                 var reason = new StringBuilder();
-                Strings.collectionToDelimitedStringWithLimit(
-                    (Iterable<String>) () -> Iterators.map(results.iterator(), t -> t.sourceIndexName() + "->" + t.rolloverIndexName()),
-                    ",",
-                    "bulk rollover [",
-                    "]",
-                    1024,
-                    reason
-                );
+                Strings.collectionToDelimitedStringWithLimit(results, ",", "bulk rollover [", "]", 1024, reason);
                 try (var ignored = batchExecutionContext.dropHeadersContext()) {
                     state = allocationService.reroute(state, reason.toString(), listener.reroute());
                 }
@@ -514,7 +506,7 @@ public ClusterState execute(BatchExecutionContext<RolloverTask> batchExecutionCo
 
         public ClusterState executeTask(
             ClusterState currentState,
-            List<MetadataRolloverService.RolloverResult> results,
+            ArrayList<String> results,
             TaskContext<RolloverTask> rolloverTaskContext,
             AllocationActionMultiListener<RolloverResponse> allocationActionMultiListener
         ) throws Exception {
@@ -586,7 +578,7 @@ public ClusterState executeTask(
                     rolloverTask.autoShardingResult(),
                     targetFailureStore
                 );
-                results.add(rolloverResult);
+                results.add(rolloverResult.sourceIndexName() + "->" + rolloverResult.rolloverIndexName());
                 logger.trace("rollover result [{}]", rolloverResult);
 
                 final var rolloverIndexName = rolloverResult.rolloverIndexName();