Centralized logging (ELK stack) in Microservices - System Design Exercise

Design: Centralized Logging System using ELK Stack

Includes log collection, storage, search, and visualization. Excludes log generation and microservice internal logging implementation.

Functional Requirements

FR1: Collect logs from multiple microservices in real-time

FR2: Store logs centrally for easy search and analysis

FR3: Provide a dashboard for monitoring logs with filtering and alerting

FR4: Support log retention for at least 30 days

FR5: Handle at least 10,000 log events per second

FR6: Ensure logs are searchable with p99 query latency under 200ms

Non-Functional Requirements

NFR1: System must be highly available with 99.9% uptime

NFR2: Logs must be securely transmitted and stored

NFR3: The system should scale horizontally as log volume grows

NFR4: Minimal impact on microservices performance when sending logs

Think Before You Design

Questions to Ask

❓ Question 1

❓ Question 2

❓ Question 3

❓ Question 4

❓ Question 5

Key Components

Log shippers (e.g., Filebeat, Logstash)

Message queue or buffer for log ingestion

Elasticsearch cluster for storage and search

Kibana for visualization and dashboards

Security components like TLS and authentication

Design Patterns

Log aggregation

Event streaming with buffering

Indexing and search optimization

Horizontal scaling and sharding

Data retention and archival

Reference Architecture

Microservices --> Filebeat/Logstash --> Kafka (buffer) --> Elasticsearch Cluster --> Kibana Dashboard
                    |                     |                     |                     |
                    |                     |                     |                     |
                    +---------------------+---------------------+---------------------+

Components

Microservices

Any microservice framework

Generate application logs in structured format

Filebeat / Logstash

Elastic Beats / Logstash

Collect and forward logs from microservices reliably

Kafka

Apache Kafka

Buffer logs to handle spikes and decouple producers from consumers

Elasticsearch Cluster

Elasticsearch

Store, index, and provide fast search over logs

Kibana

Kibana

Visualize logs, create dashboards, and set alerts

Request Flow

1. Microservices generate logs and write to local files or stdout.

2. Filebeat agents installed on microservice hosts read logs and forward them to Kafka or Logstash.

3. Kafka buffers incoming logs to handle bursts and ensure durability.

4. Logstash consumes logs from Kafka, processes and transforms them if needed, then sends to Elasticsearch.

5. Elasticsearch indexes logs for fast search and stores them with retention policies.

6. Kibana connects to Elasticsearch to provide dashboards and alerting interfaces for users.

Database Schema

Elasticsearch stores logs as documents with fields: timestamp, service_name, log_level, message, trace_id, host, and custom tags. Indexes are created on timestamp and service_name for efficient querying.

Scaling Discussion

Bottlenecks

Log ingestion rate exceeding Kafka or Logstash capacity

Elasticsearch cluster storage and query performance limits

Network bandwidth between microservices and log shippers

Kibana dashboard performance with large datasets

Solutions

Scale Kafka brokers horizontally and partition topics for parallelism

Add Elasticsearch nodes and use sharding to distribute data

Use compression and batching in Filebeat to reduce network load

Implement index lifecycle management to archive or delete old logs

Optimize Kibana queries and use filters to limit data volume

Interview Tips

Time: Spend 10 minutes understanding requirements and clarifying scale. Use 20 minutes to design components and data flow. Reserve 10 minutes to discuss scaling and trade-offs. Use last 5 minutes for questions.

Explain how logs flow from microservices to Elasticsearch

Discuss buffering with Kafka to handle spikes

Highlight importance of indexing and search optimization

Mention security and data retention considerations

Describe scaling strategies for each component