Blog Post

How to Propagate OpenTelemetry Trace Headers Over AWS Kinesis: Part 1

In this short blog series, we discuss trace header propagation in AWS Kinesis, its challenges, and solutions for embedding and extracting trace context in records.

Welcome to our series on navigating the complexities of trace header propagation with OpenTelemetry in AWS Kinesis.

In this 3-part exploration, we'll dive into the critical role of trace headers in distributed systems, discuss the unique challenges presented by AWS Kinesis, and explore innovative solutions that keep your data tracking robust and consistent.  

Understanding trace header propagation

“Trace Header” propagation in the OpenTelemetry world refers to the mechanism by which trace information, including identifiers and other metadata related to a specific request, is carried across service boundaries in a distributed system. This propagation is crucial for maintaining a continuous trace context across different system components.  

What are trace headers?

Trace headers are pieces of information, such as trace ID, span ID, and trace flags, embedded in the headers of HTTP requests or messaging protocols. The trace ID represents the overall journey of a request across the system, while the span ID points to specific operations or service calls within that journey.

How are trace headers propagated?

OpenTelemetry specifies how these trace headers are forwarded from one service to another. When a service receives a request, it extracts these headers, uses them to record tracing information, and includes them in any outgoing requests to downstream services. This ensures that each service in the path of a request contributes to a unified trace.

The role of cross-service tracing

By propagating trace headers, OpenTelemetry enables cross-service tracing. This means you can trace a request from its origin through all the services it traverses until its completion. It provides a cohesive view of a request’s path through a distributed system.

Interoperability and standardization in trace propagation

OpenTelemetry adheres to standardized formats for trace context propagation, such as W3C Trace Context. This standardization ensures compatibility and interoperability between different tracing tools and platforms, making integrating and switching between different observability solutions easier.

The significance of trace propagation in distributed systems

In microservices or distributed architectures, a single user request might involve multiple services. Trace header propagation allows developers and operators to visualize the entire path of a request, making it easier to diagnose issues, understand service dependencies, and optimize performance.

Understanding the challenges of trace header propagation in AWS Kinesis

In today’s rapidly evolving digital landscape, tracking and monitoring distributed applications is crucial for ensuring performance and reliability. Trace header propagation is essential to this process as it helps to track and optimize the flow of requests across various services. While HTTP-based communications facilitate straightforward trace propagation through headers, the scenario is markedly different for services like AWS Kinesis, where the absence of a separate metadata layer presents unique challenges.  

In HTTP communication, trace headers are typically added to the HTTP headers, separate from the body of the request. This allows trace information to be propagated without modifying the actual content of the message. Tools like OpenTelemetry provide mechanisms to inject and extract these headers automatically as part of the standard HTTP request and response cycle. This approach is non-intrusive and does not affect the integrity or format of the message body.

However, unlike HTTP, specific services, such as AWS Kinesis, do not inherently support a separate metadata or header section for each record where trace context can be seamlessly added. For example, in AWS Kinesis, each record is a blob of data, and there is no built-in mechanism to attach headers or metadata directly to individual records. As a result, to propagate trace information, you often need to modify the actual record to include this information.

This blog series will explore the fundamental hurdles encountered in trace header propagation through AWS Kinesis. More importantly, we will discuss innovative and practical solutions to overcome these challenges. These solutions aim to enhance the efficiency of trace data transmission and improve the effectiveness of monitoring processes in distributed systems.  

What is the current solution to propagate trace headers over the AWS Kinesis service?

In AWS Kinesis, propagating trace headers between producers and consumers involves embedding trace information into the records sent by producers and extracting it on the consumer side. This process allows you to trace the journey of data through Kinesis streams and is crucial for monitoring and troubleshooting in distributed systems.

Here's how you can achieve this:

Embedding trace headers in Kinesis records (producer side)

  • When a producer application sends a record to a Kinesis stream, it should include trace context information within the record. This can be done by adding trace headers (like trace ID and span ID) to the record's data.
  • The trace context can be obtained from the OpenTelemetry SDK or similar tracing tools integrated into the producer application.
  • This trace context should conform to a standard format (e.g., W3C Trace Context) to ensure compatibility and ease of use.    

Sending records to AWS Kinesis

  • Use the AWS SDK to send the record to a Kinesis stream. Ensure that the trace context information is included in each record payload.
  • It’s important to maintain the integrity and readability of the trace context so that the consumer can correctly interpret it.

Extracting trace headers in Kinesis records (consumer side)

  • On the consumer side, applications or services reading from the Kinesis stream should extract the trace context information from each record.
  • This involves parsing the record data to retrieve the trace headers.
  • Once extracted, this trace context should be used to continue the trace, linking the consumer's processing activities with the trace initiated by the producer.

Continuing the trace

  • After extracting the trace context, the consumer application should use it to annotate its own processing activities, creating spans that are logically connected to those created by the producer.
  • This continued tracing helps in visualizing the entire lifecycle of data as it moves from the producer through Kinesis and to the consumer.

What are the challenges with modifying records to propagate trace headers over AWS Kinesis?

In AWS Kinesis, modifying the original content of a record to include trace headers for propagation can lead to several challenges and potential problems.  

The necessity to alter the original message to include trace headers introduces issues such as maintaining data integrity, managing increased message size, and ensuring compatibility with downstream systems expecting a specific data format.

Additionally, it adds complexity to the data processing logic, as producers need to embed the trace context into the data, and consumers must extract and interpret it correctly.

Here are some issues that might arise:

  1. Data integrity concerns - Altering the original record to include trace headers can compromise the integrity of the data. This is particularly critical if the data is being used for sensitive or regulatory-compliant processes where tampering or changes could have serious implications.
  2. Schema and format compatibility - If the data schema is strict or predefined (e.g., in a data pipeline expecting a specific format), adding additional trace headers might cause compatibility issues with downstream systems expecting a certain structure.
  3. Complexity in Record Processing - Including trace headers within the record adds complexity to the processing logic on both the producer and consumer sides. Producers need to embed trace information correctly, and consumers must parse and extract this information, which adds additional processing steps.

Part 1 Conclusion

As we’ve seen, incorporating trace headers directly into records introduces a layer of complexity to AWS Kinesis operations, challenging both producers and consumers in the system. While we navigate these complexities, it's clear that an effective solution must simplify this process without sacrificing the granularity of data tracing.

In the next installment of our series, we'll delve into the innovative approaches that address these challenges, offering a more streamlined method for trace header propagation.

Application Experience
This is some text inside of a div block.

You might also like

Blog post

How to Propagate OpenTelemetry Trace Headers Over AWS Kinesis: Part 3

Blog post

How to Propagate OpenTelemetry Trace Headers Over AWS Kinesis: Part 2

Blog post

How to Propagate OpenTelemetry Trace Headers Over AWS Kinesis: Part 1