Observability in Cloud‑Native Applications with Elastic Stack: A Four‑Step Approach

At the 2nd Cloud Native Technology Practice Summit (CNBPS 2019), Elastic China evangelist Liu Zheng presented "Not Just Logs, Elastic Stack Gives You More" focusing on building observability for cloud‑native applications.

1. Characteristics and Operational Challenges of Cloud‑Native Apps – Cloud‑native workloads are characterized by microservice architectures, containerized infrastructure, and continuous delivery pipelines, which increase deployment complexity, scaling demands, and the need for rapid rollback and fault isolation.

2. The Essence of Observability – Observability goes beyond monitoring by answering why a component is not working, requiring logs, metrics, and tracing signals across all layers of the system.

3. Four‑Step Method to Build Observability – The speaker proposes a layered approach: Health Checks (Level 0), Metrics (Level 1), Logs (Level 2), and Application Tracing/APM (Level 3). Each step adds depth to the insight into system behavior.

Level 0 – Health Checks – Elastic Stack provides a red‑green status system and supports broadcast, registry (etcd/ZooKeeper), or expose modes for service health reporting.

Level 1 – Metrics – Collect system, application, and business metrics via agents (Beats) or pull‑based scrapers (e.g., Prometheus), store them in Elasticsearch, and enable alerting and aggregation.

Level 2 – Logs – Centralize log ingestion with Logstash and Beats, ensure full‑text search, correlation, and retention for both request and error logs.

Level 3 – Tracing (APM) – Use Elastic APM to capture distributed traces, pinpoint slow operations, and correlate with logs and metrics for end‑to‑end debugging.

The speaker emphasizes that a reliable, scalable storage backend (Elasticsearch) and low‑cost tooling are key to a successful observability implementation, and encourages using Elastic Stack’s integrated features to achieve a complete monitoring solution.