Setting Up Docker Containers for Tracing with Phoenix Collector

abhijit h. · 2024-08-16T17:19:57.781Z

Hi, I am having my agent deployed as a docker. And I have 2 instances of my agent, and I want to export the traces from the 2 docker container. Is there a example on how to setup my docker containers so that it can send the traces to the pheonix collector ? Also, I want to deploy pheonix as a standalone docker and expose the Endpoint. Please, can someone point me to a example?

Setting Up Docker Containers for Tracing with Phoenix Collector | Arize AI Community

2 comments

· Sorted by Oldest

RunLLM
APP
·
To set up your Docker containers to send traces to a Phoenix collector and deploy Phoenix as a standalone Docker container, follow these steps: 1. Deploy Phoenix as a Standalone Docker Container: - Pull the Phoenix Docker image and run it, exposing the necessary ports:
```LBKS - This command will start the Phoenix server, exposing port 6006 for the UI and OTLP HTTP collector, and port 4317 for the OTLP gRPC collector ([source](https://docs.arize.com/phoenix/deployment/deploying-phoenix#03132250edec410c927a4cdc4fff90fe)).LBKS2. **Configure Your Agent Docker Containers to Send Traces**: - Ensure your agent containers are configured to send traces to the Phoenix collector endpoint. Here is an example of how to set up the environment variables and instrumentation in your agent's Dockerfile or entrypoint script: ```LBKS import os from opentelemetry import trace as trace_api from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter from opentelemetry.sdk import trace as trace_sdk from opentelemetry.sdk.resources import Resource from opentelemetry.sdk.trace.export import SimpleSpanProcessor from openinference.instrumentation.langchain import LangChainInstrumentorLBKS # Set the Phoenix collector endpoint os.environ['PHOENIX_COLLECTOR_ENDPOINT'] = 'http://<phoenix_host> :6006/v1/traces'LBKS # Initialize the tracer provider resource = Resource(attributes={}) tracer_provider = trace_sdk.TracerProvider(resource=resource) span_exporter = OTLPSpanExporter(endpoint=os.environ['PHOENIX_COLLECTOR_ENDPOINT']) tracer_provider.add_span_processor(SimpleSpanProcessor(span_exporter)) trace_api.set_tracer_provider(tracer_provider=tracer_provider)LBKS # Instrument LangChain LangChainInstrumentor().instrument() ```LBKSLBKS3. **Docker Compose Example**: - You can use Docker Compose to manage multiple containers and ensure they are correctly networked. Here is an example `docker-compose.yml` file: ```LBKS version: '3.8' services: phoenix: image: arizephoenix/phoenix:latest ports: - "6006:6006" # UI and OTLP HTTP collector - "4317:4317" # OTLP gRPC collectorLBKS agent1: build: ./agent1 environment: - PHOENIX_COLLECTOR_ENDPOINT=[http://phoenix:6006/v1/traces](http://phoenix:6006/v1/traces) depends_on: - phoenixLBKS agent2: build: ./agent2 environment: - PHOENIX_COLLECTOR_ENDPOINT=[http://phoenix:6006/v1/traces](http://phoenix:6006/v1/traces) depends_on: - phoenix
RunLLM
APP
·
- This setup ensures that both agent containers send their traces to the Phoenix collector running in the phoenix service (source).Would you like more details on any specific part of this setup?Sources: - Phoenix Deployment Documentation - Slack Discussion on Phoenix Setup
💡 Hint: Mention RunLLM in the thread for followups.

RunLLM
APP
·
To set up your Docker containers to send traces to a Phoenix collector and deploy Phoenix as a standalone Docker container, follow these steps: 1. Deploy Phoenix as a Standalone Docker Container: - Pull the Phoenix Docker image and run it, exposing the necessary ports:
```LBKS - This command will start the Phoenix server, exposing port 6006 for the UI and OTLP HTTP collector, and port 4317 for the OTLP gRPC collector ([source](https://docs.arize.com/phoenix/deployment/deploying-phoenix#03132250edec410c927a4cdc4fff90fe)).LBKS2. **Configure Your Agent Docker Containers to Send Traces**: - Ensure your agent containers are configured to send traces to the Phoenix collector endpoint. Here is an example of how to set up the environment variables and instrumentation in your agent's Dockerfile or entrypoint script: ```LBKS import os from opentelemetry import trace as trace_api from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter from opentelemetry.sdk import trace as trace_sdk from opentelemetry.sdk.resources import Resource from opentelemetry.sdk.trace.export import SimpleSpanProcessor from openinference.instrumentation.langchain import LangChainInstrumentorLBKS # Set the Phoenix collector endpoint os.environ['PHOENIX_COLLECTOR_ENDPOINT'] = 'http://<phoenix_host> :6006/v1/traces'LBKS # Initialize the tracer provider resource = Resource(attributes={}) tracer_provider = trace_sdk.TracerProvider(resource=resource) span_exporter = OTLPSpanExporter(endpoint=os.environ['PHOENIX_COLLECTOR_ENDPOINT']) tracer_provider.add_span_processor(SimpleSpanProcessor(span_exporter)) trace_api.set_tracer_provider(tracer_provider=tracer_provider)LBKS # Instrument LangChain LangChainInstrumentor().instrument() ```LBKSLBKS3. **Docker Compose Example**: - You can use Docker Compose to manage multiple containers and ensure they are correctly networked. Here is an example `docker-compose.yml` file: ```LBKS version: '3.8' services: phoenix: image: arizephoenix/phoenix:latest ports: - "6006:6006" # UI and OTLP HTTP collector - "4317:4317" # OTLP gRPC collectorLBKS agent1: build: ./agent1 environment: - PHOENIX_COLLECTOR_ENDPOINT=[http://phoenix:6006/v1/traces](http://phoenix:6006/v1/traces) depends_on: - phoenixLBKS agent2: build: ./agent2 environment: - PHOENIX_COLLECTOR_ENDPOINT=[http://phoenix:6006/v1/traces](http://phoenix:6006/v1/traces) depends_on: - phoenix
RunLLM
APP
·
- This setup ensures that both agent containers send their traces to the Phoenix collector running in the phoenix service (source).Would you like more details on any specific part of this setup?Sources: - Phoenix Deployment Documentation - Slack Discussion on Phoenix Setup
💡 Hint: Mention RunLLM in the thread for followups.