We faced Issues during Migration to Isolated Worker Process

Question

We faced Issues during Migration to Isolated Worker Process

Sefa Teyek 0

Dear Microsoft Support Team,

We are in the process of migrating our application from the in-process model to Azure Functions Isolated Worker, following all official Microsoft guidelines.

While we have updated the configuration as suggested, we are currently facing two major issues:

[Singleton] Attribute No Longer Supported

Our application logic heavily relies on the [Singleton] attribute, which is no longer supported in the isolated worker model.

This poses a significant challenge in terms of:

Ensuring single-instance execution for certain critical operations

Preserving the same behavior without major architectural overhaul

We are exploring alternatives like:

Service Bus sessions for message ordering and instance locking

Distributed locking

We’d appreciate it if you could provide Microsoft-recommended patterns or best practices to safely replace the [Singleton] attribute and maintain the original singleton behavior within the isolated worker model.

Logging Gaps in Isolated Worker Model

We are also experiencing serious gaps in telemetry (Application Insights) since the migration:

operation_name is missing for many traces, exceptions, dependencies, and custom events

Custom dimensions are intermittently dropped or missing entirely

These issues were not present in the in-process model

This telemetry is critical to our business for monitoring, debugging, and custom workbook dashboards.

Could you advise on:

How to preserve full log fidelity (including operation_name and custom dimensions) in the isolated worker model?

Whether there's a recommended approach or configuration for bridging this telemetry gap?

We’d really appreciate your guidance on these issues, as they’re essential for ensuring feature parity and observability in the new hosting model.

Shireesha Eeraboina 3,435 Reputation points Microsoft External Staff Moderator

2025-04-17T07:32:30.76+00:00
Hi Sefa Teyek,

For the issues you're facing with migrating to the Azure Functions Isolated Worker.

[Singleton] Attribute:

Since the [Singleton] attribute isn't directly supported, you'll need to use external mechanisms to ensure single-instance execution. The Microsoft-recommended approaches are distributed locking using Azure Blob Storage leases or Azure Cache for Redis. Service Bus sessions can also work if your functions are triggered by Service Bus.

Please refer these documentation:

Service Bus Sessions

Logging Gaps:

To maintain full log fidelity in the Isolated Worker model, you need to explicitly manage telemetry. Ensure you're using the Microsoft.ApplicationInsights.WorkerService package and manually set the operation_name and custom dimensions in your code.

To help you better understand, kindly refer to the documentations below:

Singleton alternatives: https://learn.microsoft.com/azure/architecture/patterns/leader-election

Logging in Isolated model: https://learn.microsoft.com/azure/azure-functions/dotnet-isolated-process-guide#logging-and-telemetry

For more clarifications and details, please review these threads for a better understanding.

https://learn.microsoft.com/en-us/answers/questions/1420017/migration-from-azure-function-in-process-model-to

https://learn.microsoft.com/en-us/answers/questions/1661100/migrate-function-app-from-in-process-to-isolated-p

I hope this addresses your query. Please let me know if you need any further assistance or clarification.
Shireesha Eeraboina 3,435 Reputation points Microsoft External Staff Moderator

2025-04-18T05:05:26.8133333+00:00

Hello Sefa Teyek,

Following up to see if you had a chance to review the above response. If you have any further updates on this issue, please feel free to post them here.
Sefa Teyek 0 Reputation points

2025-04-18T08:29:21.8033333+00:00

Hello @Shireesha Eeraboina

We are currently reviewing the Microsoft documentation and applying all relevant updates to our codebase.

If the issue with the operation name in the logging persists, I will provide additional details, including example code snippets, library versions, and AppInsights logs for further analysis.
Sai Prabhu Naveen Parimi 2,425 Reputation points Microsoft External Staff Moderator

2025-04-22T08:52:05.32+00:00

@Sefa Teyek

Just checking in — were you able to gather the additional details (like code snippets, library versions, or App Insights logs) for the operation_name issue? Happy to take a closer look and help further if needed.

1 answer

Your answer

Shireesha Eeraboina 3,435 Reputation points Microsoft External Staff Moderator

2025-04-17T07:32:30.76+00:00

Hi Sefa Teyek,

For the issues you're facing with migrating to the Azure Functions Isolated Worker.

[Singleton] Attribute:

Since the [Singleton] attribute isn't directly supported, you'll need to use external mechanisms to ensure single-instance execution. The Microsoft-recommended approaches are distributed locking using Azure Blob Storage leases or Azure Cache for Redis. Service Bus sessions can also work if your functions are triggered by Service Bus.

Please refer these documentation:

Service Bus Sessions

Logging Gaps:

To maintain full log fidelity in the Isolated Worker model, you need to explicitly manage telemetry. Ensure you're using the Microsoft.ApplicationInsights.WorkerService package and manually set the operation_name and custom dimensions in your code.

To help you better understand, kindly refer to the documentations below:

Singleton alternatives: https://learn.microsoft.com/azure/architecture/patterns/leader-election

Logging in Isolated model: https://learn.microsoft.com/azure/azure-functions/dotnet-isolated-process-guide#logging-and-telemetry

For more clarifications and details, please review these threads for a better understanding.

https://learn.microsoft.com/en-us/answers/questions/1420017/migration-from-azure-function-in-process-model-to

https://learn.microsoft.com/en-us/answers/questions/1661100/migrate-function-app-from-in-process-to-isolated-p

I hope this addresses your query. Please let me know if you need any further assistance or clarification.
Shireesha Eeraboina 3,435 Reputation points Microsoft External Staff Moderator

2025-04-18T05:05:26.8133333+00:00

Hello Sefa Teyek,

Following up to see if you had a chance to review the above response. If you have any further updates on this issue, please feel free to post them here.
Sefa Teyek 0 Reputation points

2025-04-18T08:29:21.8033333+00:00

Hello @Shireesha Eeraboina

We are currently reviewing the Microsoft documentation and applying all relevant updates to our codebase.

If the issue with the operation name in the logging persists, I will provide additional details, including example code snippets, library versions, and AppInsights logs for further analysis.
Sai Prabhu Naveen Parimi 2,425 Reputation points Microsoft External Staff Moderator

2025-04-22T08:52:05.32+00:00

@Sefa Teyek

Just checking in — were you able to gather the additional details (like code snippets, library versions, or App Insights logs) for the operation_name issue? Happy to take a closer look and help further if needed.

Answer 1

Hello!

@Sai Prabhu Naveen Parimi we are working with @Sefa Teyek and trying to resolve above issues together.

So we've tried to implement a custom approach to enforce writing operation_name in our appinsights logs.

We added custom middleware to start new operation and explicitly add "CustomOperationName" for current activity on every function invocation.

public class TelemetryCorrelationMiddleware : IFunctionsWorkerMiddleware
    {
        private readonly TelemetryClient _telemetryClient;

        public TelemetryCorrelationMiddleware(TelemetryClient telemetryClient)
        {
            _telemetryClient = telemetryClient.VerifyNotNull(nameof(telemetryClient));
        }

        public async Task Invoke(FunctionContext context, FunctionExecutionDelegate next)
        {
            var functionName = context.FunctionDefinition.Name;

            using var activity = new Activity(functionName);

            activity.AddTag("CustomOperationName", functionName);
            activity.AddBaggage("CustomOperationName", functionName);

            using (var operation = _telemetryClient.StartOperation<RequestTelemetry>(activity))
            {
                await next(context);
            }
        }
    }

on the other side we've added custom TelemetryInitializer where we manually check for "CustomOperationName" and try to set it explicitly if possible

public class CustomTelemetryInitializer : ITelemetryInitializer
    {
        private const string CustomOperationName = "CustomOperationName";
        public void Initialize(ITelemetry telemetry)
        {
            if (telemetry is TraceTelemetry traceTelemetry && string.IsNullOrEmpty(traceTelemetry.Context.Operation.Name))
            {
                if (traceTelemetry.Properties.TryGetValue(CustomOperationName, out var operationName))
                {
                    traceTelemetry.Context.Operation.Name = operationName;
                }
            }

            if (telemetry is DependencyTelemetry dependencyTelemetry && string.IsNullOrEmpty(dependencyTelemetry.Context.Operation.Name))
            {
                if (dependencyTelemetry.Properties.TryGetValue(CustomOperationName, out var operationName))
                {
                    dependencyTelemetry.Context.Operation.Name = operationName;
                }
            }

            if (telemetry is EventTelemetry eventTelemetry && string.IsNullOrEmpty(eventTelemetry.Context.Operation.Name))
            {
                if (eventTelemetry.Properties.TryGetValue(CustomOperationName, out var operationName))
                {
                    eventTelemetry.Context.Operation.Name = operationName;
                }
            }

            if (telemetry is ExceptionTelemetry exceptionTelemetry && string.IsNullOrEmpty(exceptionTelemetry.Context.Operation.Name))
            {
                if (exceptionTelemetry.Properties.TryGetValue(CustomOperationName, out var operationName))
                {
                    exceptionTelemetry.Context.Operation.Name = operationName;
                }
            }
        }
    }

As a result:

Traces -> inside functions call are written with operation name
Dependencies -> OK
Custom Events -> OK
Exceptions -> OK
Requests -> No operation name but there is a column "Name" that basically has the same value.

As for [Singleton] we will try to implement solution to this issue in nearest future. I will share a feedback here.

We are considering two major approaches:

Message Sessions
Blob lease distributed lock

Andrew Baran 0 Reputation points

2025-04-22T10:43:37.7733333+00:00

If you have any concerns with this approach please feel free to comment on it.
Sai Prabhu Naveen Parimi 2,425 Reputation points Microsoft External Staff Moderator

2025-04-23T06:06:07.5866667+00:00

@Andrew Baran

Thanks again for sharing the telemetry workaround — it’s really useful and could help others going through the same migration.

When you finalize your [Singleton] replacement using Blob lease or Service Bus sessions, it would be awesome if you could post the details here as well. That way, anyone referring to this thread can benefit directly from your solution.

In case it’s helpful, here are some relevant docs:

Leader Election pattern (for distributed locks)

Azure Service Bus sessions
Sai Prabhu Naveen Parimi 2,425 Reputation points Microsoft External Staff Moderator

2025-04-24T08:31:44.8666667+00:00

@Andrew Baran

Just wanted to check in to see if you have any questions or need further clarification on the approach shared earlier.
Andrew Baran 0 Reputation points

2025-04-25T09:47:02.9+00:00

@Sai Prabhu Naveen Parimi thank you for your feedback!

At this point we don't have any specific issue with logs.

We will proceed to test it internally.

Share via

We faced Issues during Migration to Isolated Worker Process

1 answer

Your answer