Skip to content
Open
Show file tree
Hide file tree
Changes from 6 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
227 changes: 0 additions & 227 deletions Test/DurableTask.AzureStorage.Tests/OrchestrationSessionTests.cs

This file was deleted.

22 changes: 15 additions & 7 deletions src/DurableTask.AzureStorage/AzureStorageOrchestrationService.cs
Original file line number Diff line number Diff line change
Expand Up @@ -553,7 +553,7 @@ internal async Task OnIntentLeaseAquiredAsync(BlobPartitionLease lease)
{
var controlQueue = new ControlQueue(this.azureStorageClient, lease.PartitionId, this.messageManager);
await controlQueue.CreateIfNotExistsAsync();
this.orchestrationSessionManager.ResumeListeningIfOwnQueue(lease.PartitionId, controlQueue, this.shutdownSource.Token);
await this.orchestrationSessionManager.ResumeListeningIfOwnQueue(lease.PartitionId, controlQueue, this.shutdownSource.Token);
}

internal Task OnIntentLeaseReleasedAsync(BlobPartitionLease lease, CloseReason reason)
Expand All @@ -572,21 +572,20 @@ internal async Task OnOwnershipLeaseAquiredAsync(BlobPartitionLease lease)
this.allControlQueues[lease.PartitionId] = controlQueue;
}

internal void DropLostControlQueue(TablePartitionLease partition)
internal async Task DropLostControlQueue(TablePartitionLease partition)
{
// If lease is lost but we're still dequeuing messages, remove the queue
if (this.allControlQueues.TryGetValue(partition.RowKey, out ControlQueue controlQueue) &&
this.OwnedControlQueues.Contains(controlQueue) &&
partition.CurrentOwner != this.settings.WorkerId)
{
this.orchestrationSessionManager.RemoveQueue(partition.RowKey, CloseReason.LeaseLost, nameof(DropLostControlQueue));
await this.orchestrationSessionManager.RemoveQueue(partition.RowKey, CloseReason.LeaseLost, nameof(DropLostControlQueue));
}
}

internal Task OnOwnershipLeaseReleasedAsync(BlobPartitionLease lease, CloseReason reason)
{
this.orchestrationSessionManager.RemoveQueue(lease.PartitionId, reason, "Ownership LeaseCollectionBalancer");
return Utils.CompletedTask;
return this.orchestrationSessionManager.RemoveQueue(lease.PartitionId, reason, "Ownership LeaseCollectionBalancer");
}

internal async Task OnTableLeaseAcquiredAsync(TablePartitionLease lease)
Expand Down Expand Up @@ -1249,7 +1248,13 @@ await this.CommitOutboundQueueMessages(
{
var messages = session.DeferredMessages.ToList();
session.DeferredMessages.Clear();
this.orchestrationSessionManager.AddMessageToPendingOrchestration(session.ControlQueue, messages, session.TraceActivityId, CancellationToken.None);
IReadOnlyList<MessageData> messagesToAbandon = this.orchestrationSessionManager.AddMessageToPendingOrchestration(
session.ControlQueue,
messages,
session.TraceActivityId,
CancellationToken.None);

await this.orchestrationSessionManager.AbandonMessagesForDrainAsync(session.ControlQueue, messagesToAbandon);
}
}
// Handle the case where the 'ETag' has changed, which implies another worker has taken over this work item while
Expand Down Expand Up @@ -1524,8 +1529,11 @@ async Task ReleaseSessionAsync(string instanceId)
if (this.orchestrationSessionManager.TryReleaseSession(
instanceId,
this.shutdownSource.Token,
out OrchestrationSession session))
out OrchestrationSession session,
out IReadOnlyList<MessageData> messagesToAbandon))
{
await this.orchestrationSessionManager.AbandonMessagesForDrainAsync(session.ControlQueue, messagesToAbandon);

// Some messages may need to be discarded
await session.DiscardedMessages.ParallelForEachAsync(
this.settings.MaxStorageOperationConcurrency,
Expand Down
45 changes: 45 additions & 0 deletions src/DurableTask.AzureStorage/Messaging/ControlQueue.cs
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,7 @@ namespace DurableTask.AzureStorage.Messaging
using DurableTask.AzureStorage.Monitoring;
using DurableTask.AzureStorage.Partitioning;
using DurableTask.AzureStorage.Storage;
using DurableTask.Core;

class ControlQueue : TaskHubQueue, IDisposable
{
Expand Down Expand Up @@ -209,6 +210,50 @@ public override Task AbandonMessageAsync(MessageData message, SessionBase? sessi
return base.AbandonMessageAsync(message, session);
}

/// <summary>
/// Abandons a message with zero visibility timeout so it becomes immediately visible
/// for another partition owner to pick up. This is used during drain to avoid stranding
/// messages that were dequeued but not yet promoted to active sessions.
/// </summary>
public async Task AbandonMessageForDrainAsync(MessageData message)
{
this.stats.PendingOrchestratorMessages.TryRemove(message.OriginalQueueMessage.MessageId, out _);

QueueMessage queueMessage = message.OriginalQueueMessage;
TaskMessage taskMessage = message.TaskMessage;
OrchestrationInstance instance = taskMessage.OrchestrationInstance;

this.settings.Logger.AbandoningMessage(
this.storageAccountName,
this.settings.TaskHubName,
taskMessage.Event.EventType.ToString(),
Utils.GetTaskEventId(taskMessage.Event),
queueMessage.MessageId,
instance.InstanceId,
instance.ExecutionId,
this.storageQueue.Name,
message.SequenceNumber,
queueMessage.PopReceipt,
visibilityTimeoutSeconds: 0);

try
{
await this.storageQueue.UpdateMessageAsync(
queueMessage,
TimeSpan.Zero,
clientRequestId: null);
}
catch (DurableTaskStorageException e)
{
this.settings.Logger.PartitionManagerWarning(
this.storageAccountName,
this.settings.TaskHubName,
this.settings.WorkerId,
this.Name,
$"Failed to abandon message {queueMessage.MessageId} during drain: {e.Message}");
}
}

public override Task DeleteMessageAsync(MessageData message, SessionBase? session = null)
{
this.stats.PendingOrchestratorMessages.TryRemove(message.OriginalQueueMessage.MessageId, out _);
Expand Down
Loading
Loading