Logging and ranting
summary
Company standardizes on a central JSON-over-stdout contract so containers and nginx emit structured logs, routed into an ELK stack; dynamic mapping causes thousands of fields so templates are enforced. An entry Kafka buffer absorbs bursts — tested up to 100k msg/s, average ingestion ~25k msg/s — with 24‑hour indexing and retention to limit cost (example: 200 GB/hour → ~$12k/month on AWS). Adds a trace ID propagated ad hoc across services for request-level tracing; PHP stdout limits force a UNIX-socket fallback.
AI-written summary · about
chapters (12)
- 0:00Introduction and context
- 2:07Logging purpose and methods
- 3:38Structured logging strategy
- 7:02The ELK stack implementation
- 9:13Challenges with log structure
- 12:25Data mapping and visualization
- 15:44Log volume and cost
- 18:39Tracing and distributed systems
- 22:46Defining events and metrics
- 26:47Anomalies and automated reactions
- 29:30Summary of logging principles
- 30:37Q&A and technical architecture
related talks
31mLogging at Scale, Billion+ messages, 100K msg/sec
21mLogging at scale using Graylog - Billion+ messages, 100K req/sec
39mFast log management for your infrastructure
20mFast log management for your apps
31mThe Long and Winding Road to Application Logging [A]
46mBuilding a scalable logging platform
44mCentralized Logging Patterns
01h 00mApplication Logging & Logstash