OpenTelemetry Export

Because the 6.x iOS SDK is built on OpenTelemetry, it has the ability to export OpenTelemetry signals directly from the mobile code level, without any of the telemetry hitting our backend.

To send traces and logs from the SDK to your collector or vendor of choice, you will need to configure the SDK with an exporter capable of sending OTel signals to that destination. This will be done in the creation of the Embrace.Options, by adding an OpenTelemetryExport.

Direct Exporters

Some collectors have built or presently support direct export of traces or logs in Swift. In theory, any implementation of SpanExporter or LogRecordExporter that can point to the location of the collector should be able to send, respectively, spans and logs.

The OpenTelemetry-Swift repository lists publicly-available exporters that can be added directly to your Embrace configuration. For example, here is an SDK configuration that adds a Jaeger exporter for traces:

try? Embrace
    .setup(
        options: Embrace.Options(
            appId: "AppID",
            logLevel: .debug,
            export: OpenTelemetryExport(
                spanExporter: JaegerSpanExporter(
                    serviceName: "jaegerServiceName",
                    collectorAddress: "jaegerCollectorAddress"
                )
            )
        )
    )
    .start()

OTLP Export through HTTP or gRPC

The OpenTelemetry-Swift list also has OTLP HTTP and gRPC exporters for logs and spans. These can be used more flexibly than the single-service exporters like Jaeger, because vendors can provide some important keys or headers that allow you to use the protocol to export to an HTTP or gRPC address.

For example, Grafana Cloud allows you to generate a token that you can use with their OTLP traces and spans gateway. On the SDK side, you can add an OtlpHttpTraceExporter to send your spans to that Grafana account via the Grafana Cloud traces endpoint, and similarly use an OtlpHttpLogExporter to send your logs to the same account via the GC log endpoint:

let grafanaCloudTokenString = //String generated from your account
let urlConfig = URLSessionConfiguration.default
urlConfig.httpAdditionalHeaders = ["Authorization": "Basic \(grafanaCloudTokenString)"]

try? Embrace
    .setup(
        options: Embrace.Options(
            appId: "AppID",
            logLevel: .debug,
            export: OpenTelemetryExport(
                spanExporter: OtlpHttpTraceExporter(
                    endpoint: URL(string: "https://otlp-gateway-prod-us-west-0.grafana.net/otlp/v1/traces")!,
                    useSession: URLSession(configuration: urlConfig)
                ),
                logExporter: OtlpHttpLogExporter(
                    endpoint: URL(string: "https://otlp-gateway-prod-us-west-0.grafana.net/otlp/v1/logs")!,
                    useSession: URLSession(configuration: urlConfig)
                )
            )
        )
    )
    .start()

Common Use Cases

Integrating with Existing Observability Stacks

If your organization already uses an observability platform that supports OpenTelemetry, you can integrate Embrace data directly into that platform:

Send mobile app traces to the same system that monitors your backend services
Create unified dashboards that show full-stack performance
Correlate mobile issues with backend problems

Custom Exporters

For specialized environments, you can implement custom exporters that:

Send data to internal systems
Apply custom filtering or processing before export
Implement company-specific security or compliance requirements

Implementing a Custom Log Exporter

The Embrace iOS SDK supports custom log exporters through OpenTelemetry's LogRecordExporter protocol. Your custom exporter will run alongside Embrace's default storage exporter, giving you access to all log data while maintaining Embrace's core functionality.

Basic Implementation

import Foundation
import OpenTelemetrySdk

class CustomLogExporter: LogRecordExporter {

    func export(logRecords: [OpenTelemetrySdk.ReadableLogRecord], explicitTimeout: TimeInterval?) -> OpenTelemetrySdk.ExportResult {
        // Your custom export logic here
        for record in logRecords {
            // Access log data
            let message = record.body.description
            let timestamp = record.timestamp
            let attributes = record.attributes
            let severity = record.severity
            let resource = record.resource.attributes

            // Example implementations:
            // - Send to remote API
            // - Write to custom file format
            // - Forward to analytics service
            // - Apply custom filtering/transformation

            // Simple example - print to console
            print("CUSTOM EXPORTER - Message: \(message), Timestamp: \(timestamp), Severity: \(String(describing: severity))")
        }
        return .success
    }

    func forceFlush(explicitTimeout: TimeInterval?) -> OpenTelemetrySdk.ExportResult {
        // Flush any pending logs (implement your flush logic)
        print("CUSTOM EXPORTER - Force flush called")
        return .success
    }

    func shutdown(explicitTimeout: TimeInterval?) {
        // Cleanup resources when SDK shuts down
        print("CUSTOM EXPORTER - Shutdown called")
    }
}

Configuration

import EmbraceIO

// Configure your custom exporter
let customExporter = CustomLogExporter()

// Set up Embrace options
let options = Embrace.Options(
    appId: "your-app-id",
    export: OpenTelemetryExport(
        spanExporter: nil, // Optional: add custom span exporter
        logExporter: customExporter
    )
)

// Initialize Embrace
Embrace.setup(options: options)

Background Processing for Heavy Operations

For exporters that need to perform heavy processing (file I/O, network calls, data transformation), use a dedicated dispatch queue to avoid blocking the logging pipeline:

import Foundation
import OpenTelemetrySdk

class CustomLogExporter: LogRecordExporter {
    // Dedicated queue for heavy processing
    private let processingQueue = DispatchQueue(label: "com.yourapp.log-exporter", qos: .utility)

    func export(logRecords: [OpenTelemetrySdk.ReadableLogRecord], explicitTimeout: TimeInterval?) -> OpenTelemetrySdk.ExportResult {
        // Return immediately to avoid blocking the logging pipeline
        processingQueue.async {
            // Perform heavy operations here
            self.processLogsAsync(logRecords)
        }

        return .success
    }

    private func processLogsAsync(_ logRecords: [ReadableLogRecord]) {
        for record in logRecords {
            // Heavy processing examples:
            // - Save to local database
            // - Write to file system
            // - Send to remote API
            // - Complex data transformations

            self.saveToLocalFile(record)
            self.sendToAnalyticsService(record)
        }
    }

    private func saveToLocalFile(_ record: ReadableLogRecord) {
        // Example: Save to disk
        let logData = [
            "message": record.body.description,
            "timestamp": record.timestamp.timeIntervalSince1970,
            "severity": record.severity?.description ?? "unknown"
        ]

        // Write to file, database, etc.
        // This heavy I/O won't block the main logging thread
    }

    private func sendToAnalyticsService(_ record: ReadableLogRecord) {
        // Example: Network call
        // Heavy network operations happen in background
    }

    func forceFlush(explicitTimeout: TimeInterval?) -> OpenTelemetrySdk.ExportResult {
        // Wait for background processing to complete if needed
        processingQueue.sync { }
        return .success
    }

    func shutdown(explicitTimeout: TimeInterval?) {
        // Cleanup queue and resources
    }
}

Available Log Data

Each ReadableLogRecord provides:

body - Log message content
timestamp - When the log was created
severity - Log level (debug, info, warn, error, etc.)
attributes - Custom attributes attached to the log
resource - Resource information (app metadata, device info, etc.)
instrumentationScopeInfo - Information about the logging source
spanContext - Associated trace/span context if available

Example Use Cases

Send to Remote API:

func export(logRecords: [OpenTelemetrySdk.ReadableLogRecord], explicitTimeout: TimeInterval?) -> OpenTelemetrySdk.ExportResult {
    for record in logRecords {
        let logData = [
            "message": record.body.description,
            "timestamp": record.timestamp.timeIntervalSince1970,
            "severity": record.severity?.description ?? "unknown",
            "attributes": record.attributes
        ]

        // Send to your API endpoint
        sendToAPI(logData)
    }
    return .success
}

Filter and Transform Logs:

func export(logRecords: [OpenTelemetrySdk.ReadableLogRecord], explicitTimeout: TimeInterval?) -> OpenTelemetrySdk.ExportResult {
    let errorLogs = logRecords.filter { record in
        record.severity == .error || record.severity == .fatal
    }

    for record in errorLogs {
        // Process only error/fatal logs
        processErrorLog(record)
    }
    return .success
}

Important Considerations

Dual Export: Your exporter runs alongside Embrace's default exporter - both receive the same log data
Performance: Implement efficient processing to avoid blocking the logging pipeline
Error Handling: Return .failure from export methods if processing fails
Thread Safety: Ensure your exporter is thread-safe as it may be called from multiple threads
Resource Management: Clean up properly in the shutdown method

Direct Exporters​

OTLP Export through HTTP or gRPC​

Common Use Cases​

Integrating with Existing Observability Stacks​

Custom Exporters​

Implementing a Custom Log Exporter​

Basic Implementation​

Configuration​

Background Processing for Heavy Operations​

Available Log Data​

Example Use Cases​

Important Considerations​