[
https://issues.apache.org/jira/browse/IGNITE-28311?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Iurii Gerzhedovich updated IGNITE-28311:
----------------------------------------
Component/s: sql ai3
> Massive Integer boxing in index lockers
> ---------------------------------------
>
> Key: IGNITE-28311
> URL: https://issues.apache.org/jira/browse/IGNITE-28311
> Project: Ignite
> Issue Type: Bug
> Components: sql ai3
> Reporter: Denis Chudov
> Priority: Major
> Labels: ignite-3
>
> See [1]
> It creates Map<Integer, IndexLocker> using index id as a key. There is
> limited amount of index ids in the system, but every time plain int value is
> boxed causing huge heap allocations (up to 491 MiB per sample in JFR that I
> examined).
> Stack trace:
>
> {code:java}
> Stack Trace Count Percentage void java.lang.Thread.run() 1 100 % void
> java.lang.Thread.runWith(Object, Runnable) 1 100 % void
> java.util.concurrent.ThreadPoolExecutor$Worker.run() 1 100 % void
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor$Worker)
> 1 100 % void
> org.apache.ignite.internal.replicator.ReplicaService$$Lambda.0x0000000800f9c228.run()
> 1 100 % void
> org.apache.ignite.internal.replicator.ReplicaService.lambda$sendToReplicaRaw$6(CompletableFuture,
> ErrorReplicaResponse) 1 100 % boolean
> java.util.concurrent.CompletableFuture.completeExceptionally(Throwable) 1
> 100 % void java.util.concurrent.CompletableFuture.postComplete() 1 100 %
> CompletableFuture
> java.util.concurrent.CompletableFuture$UniHandle.tryFire(int) 1 100 %
> boolean java.util.concurrent.CompletableFuture.uniHandle(Object, BiFunction,
> CompletableFuture$UniHandle) 1 100 % Object
> org.apache.ignite.internal.table.distributed.storage.InternalTableImpl$$Lambda.0x0000000801167068.apply(Object,
> Object) 1 100 % CompletableFuture
> org.apache.ignite.internal.table.distributed.storage.InternalTableImpl.lambda$trackingInvoke$13(int,
> InternalTransaction, int, ReplicaRequest, PendingTxPartitionEnlistment,
> BiPredicate, Object, Throwable) 1 100 % CompletableFuture
> org.apache.ignite.internal.table.distributed.storage.InternalTableImpl.trackingInvoke(InternalTransaction,
> int, Function, boolean, PendingTxPartitionEnlistment, BiPredicate, int) 1
> 100 % CompletableFuture
> org.apache.ignite.internal.replicator.ReplicaService.invoke(String,
> ReplicaRequest) 1 100 % CompletableFuture
> org.apache.ignite.internal.replicator.ReplicaService.sendToReplica(String,
> ReplicaRequest) 1 100 % CompletableFuture
> org.apache.ignite.internal.replicator.ReplicaService.sendToReplicaRaw(String,
> ReplicaRequest) 1 100 % CompletableFuture
> org.apache.ignite.internal.network.MessagingService.invoke(String,
> NetworkMessage, long) 1 100 % CompletableFuture
> org.apache.ignite.internal.network.wrapper.JumpToExecutorByConsistentIdAfterSend.invoke(String,
> ChannelType, NetworkMessage, long) 1 100 % CompletableFuture
> org.apache.ignite.internal.network.DefaultMessagingService.invoke(String,
> ChannelType, NetworkMessage, long) 1 100 % CompletableFuture
> org.apache.ignite.internal.network.DefaultMessagingService.invoke0(InternalClusterNode,
> ChannelType, NetworkMessage, long, boolean) 1 100 % void
> org.apache.ignite.internal.network.DefaultMessagingService.sendToSelf(NetworkMessage,
> Long) 1 100 % void
> org.apache.ignite.internal.replicator.ReplicaManager$$Lambda.0x0000000800486c98.onReceived(NetworkMessage,
> InternalClusterNode, Long) 1 100 % void
> org.apache.ignite.internal.replicator.ReplicaManager.onReplicaMessageReceived(NetworkMessage,
> InternalClusterNode, Long) 1 100 % void
> org.apache.ignite.internal.replicator.ReplicaManager.handleReplicaRequest(ReplicaRequest,
> InternalClusterNode, Long) 1 100 % CompletableFuture
> org.apache.ignite.internal.replicator.ZonePartitionReplicaImpl.processRequest(ReplicaRequest,
> UUID) 1 100 % CompletableFuture
> org.apache.ignite.internal.partition.replicator.ZonePartitionReplicaListener.invoke(ReplicaRequest,
> UUID) 1 100 % CompletableFuture
> java.util.concurrent.CompletableFuture.thenCompose(Function) 1 100 %
> CompletableFuture
> java.util.concurrent.CompletableFuture.uniComposeStage(Executor, Function)
> 1 100 % Object
> org.apache.ignite.internal.partition.replicator.ZonePartitionReplicaListener$$Lambda.0x0000000801058000.apply(Object)
> 1 100 % CompletionStage
> org.apache.ignite.internal.partition.replicator.ZonePartitionReplicaListener.lambda$invoke$0(ReplicaRequest,
> UUID, ReplicaPrimacy) 1 100 % CompletableFuture
> org.apache.ignite.internal.partition.replicator.ZonePartitionReplicaListener.processRequest(ReplicaRequest,
> ReplicaPrimacy, UUID) 1 100 % CompletableFuture
> org.apache.ignite.internal.partition.replicator.ZonePartitionReplicaListener.processTableAwareRequest(ReplicaRequest,
> ReplicaPrimacy, UUID) 1 100 % CompletableFuture
> java.util.concurrent.CompletableFuture.thenCompose(Function) 1 100 %
> CompletableFuture
> java.util.concurrent.CompletableFuture.uniComposeStage(Executor, Function)
> 1 100 % Object
> org.apache.ignite.internal.partition.replicator.ZonePartitionReplicaListener$$Lambda.0x0000000801072810.apply(Object)
> 1 100 % CompletionStage
> org.apache.ignite.internal.partition.replicator.ZonePartitionReplicaListener.lambda$processTableAwareRequest$4(ReplicaRequest,
> ReplicaPrimacy, UUID, Void) 1 100 % CompletableFuture
> org.apache.ignite.internal.table.distributed.replicator.PartitionReplicaListener.process(ReplicaRequest,
> ReplicaPrimacy, UUID) 1 100 % CompletableFuture
> org.apache.ignite.internal.table.distributed.replicator.PartitionReplicaListener.processRequestInContext(ReplicaRequest,
> ReplicaPrimacy, UUID) 1 100 % CompletableFuture
> org.apache.ignite.internal.table.distributed.replicator.PartitionReplicaListener.processRequest(ReplicaRequest,
> ReplicaPrimacy) 1 100 % CompletableFuture
> org.apache.ignite.internal.table.distributed.replicator.PartitionReplicaListener.processOperationRequestWithTxOperationManagementLogic(ReplicaRequest,
> ReplicaPrimacy, HybridTimestamp) 1 100 % CompletableFuture
> org.apache.ignite.internal.table.distributed.replicator.PartitionReplicaListener.processOperationRequest(ReplicaRequest,
> ReplicaPrimacy, HybridTimestamp) 1 100 % CompletableFuture
> org.apache.ignite.internal.table.distributed.replicator.PartitionReplicaListener.appendTxCommand(UUID,
> ZonePartitionId, RequestType, boolean, Supplier) 1 100 % Object
> org.apache.ignite.internal.table.distributed.replicator.PartitionReplicaListener$$Lambda.0x000000080114b800.get()
> 1 100 % CompletableFuture
> org.apache.ignite.internal.table.distributed.replicator.PartitionReplicaListener.lambda$processOperationRequest$10(ReadWriteSingleRowPkReplicaRequest,
> ReplicaPrimacy) 1 100 % CompletableFuture
> org.apache.ignite.internal.table.distributed.replicator.PartitionReplicaListener.processSingleEntryAction(ReadWriteSingleRowPkReplicaRequest,
> long) 1 100 % CompletableFuture
> org.apache.ignite.internal.table.distributed.replicator.PartitionReplicaListener.resolveRowByPk(BinaryTuple,
> UUID, IgniteTriFunction) 1 100 % Object
> org.apache.ignite.internal.table.TableImpl$$Lambda.0x00000008010250a8.get()
> 1 100 % Map
> org.apache.ignite.internal.table.TableImpl.lambda$indexesLockers$1(int) 1
> 100 % Integer java.lang.Integer.valueOf(int) 1 100 %{code}
>
> Probably int map should be used, or field type should be changed from plain
> int to Integer.
> [1]TableImpl#indexesLockers
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
