Context Propagation
This guide explains how to propagate trace context between different execution contexts within your application using Traces.current_context
and Traces.with_context
.
Overview
The traces
library provides two complementary approaches for managing trace context:
- Local context propagation (
Traces.current_context
/Traces.with_context
): For passing context between execution contexts within the same process (threads, fibers, async tasks). - Distributed context propagation (
Traces.inject
/Traces.extract
): For transmitting context across process and service boundaries via serialization (HTTP headers, message metadata, etc.).
There is a legacy interface Traces.trace_context
and Traces.trace_context=
but you should prefer to use the new methods outlined above.
Local Context Propagation
Local context propagation involves passing trace context between different execution contexts within the same process. This is essential for maintaining trace continuity when code execution moves between threads, fibers, async tasks, or other concurrent execution contexts. Unlike distributed propagation which requires serialization over network boundaries, local propagation uses Context objects directly.
Capturing the Current Context
Use Traces.current_context
to capture the current trace context as a Context object:
current_context = Traces.current_context
# Returns a Traces::Context object or nil if no active trace
Using the Context
Use Traces.with_context(context)
to execute code within a specific trace context:
# With block (automatic restoration):
Traces.with_context(context) do
# Code runs with the specified context.
end
# Without block (permanent switch):
Traces.with_context(context)
# Context remains active.
Use Cases
Thread-Safe Context Propagation
When spawning background threads, you often want them to inherit the current trace context:
require 'traces'
# Main thread has active tracing
Traces.trace("main_operation") do
# Capture current context before spawning thread:
current_context = Traces.current_context
# Spawn background thread:
Thread.new do
# Restore context in the new thread:
Traces.with_context(current_context) do
# This thread now has the same trace context as main thread:
Traces.trace("background_work") do
perform_heavy_computation
end
end
end.join
end
Fiber-Based Async Operations
For fiber-based concurrency (like in async frameworks), context propagation ensures trace continuity:
require 'traces'
Traces.trace("main_operation") do
current_context = Traces.current_context
# Create fiber for async work:
fiber = Fiber.new do
Traces.with_context(current_context) do
# Fiber inherits the trace context:
Traces.trace("fiber_work") do
perform_async_operation
end
end
end
fiber.resume
end
Context Propagation vs. New Spans
Remember that context propagation maintains the same trace, while trace()
creates new spans:
Traces.trace("parent") do
context = Traces.current_context
Thread.new do
# This maintains the same trace context:
Traces.with_context(context) do
# This creates a NEW span within the same trace:
Traces.trace("child") do
# Child span, same trace as parent
end
end
end
end
Distributed Context Propagation
Distributed context propagation involves transmitting trace context across process and service boundaries. Unlike local propagation which works within a single process, distributed propagation requires serializing context data and transmitting it over network protocols.
Injecting Context into Headers
Use Traces.inject(headers, context = nil)
to add W3C Trace Context headers to a headers hash for transmission over network boundaries:
require 'traces'
# Capture current context:
context = Traces.current_context
headers = {'Content-Type' => 'application/json'}
# Inject trace headers:
Traces.inject(headers, context)
# headers now contains: {'Content-Type' => '...', 'traceparent' => '00-...'}
# Or use current context by default:
Traces.inject(headers) # Uses current trace context
Extracting Context from Headers
Use Traces.extract(headers)
to extract trace context from W3C headers received over the network:
# Receive headers from incoming request:
incoming_headers = request.headers
# Extract context:
context = Traces.extract(incoming_headers)
# Returns a Traces::Context object or nil if no valid context
# Use the extracted context:
if context
Traces.with_context(context) do
# Process request with distributed trace context
end
end
Use Cases
Outgoing HTTP Requests
require 'traces'
class ApiClient
def make_request(endpoint, data)
Traces.trace("api_request", attributes: {endpoint: endpoint}) do
headers = {
'content-type' => 'application/json'
}
# Add trace context to outgoing request:
Traces.inject(headers)
http_client.post(endpoint,
body: data.to_json,
headers: headers
)
end
end
end
Incoming HTTP Requests
require 'traces'
class WebController
def handle_request(request)
# Extract trace context from incoming headers:
context = Traces.extract(request.headers)
# Process request with inherited context:
if context
Traces.with_context(context) do
Traces.trace("web_request", attributes: {
path: request.path,
method: request.method
}) do
process_business_logic
end
end
else
Traces.trace("web_request", attributes: {
path: request.path,
method: request.method
}) do
process_business_logic
end
end
end
end