แนวคิดการบันทึกล็อกแบบรวมศูนย์

Imagine you have many applications running on different servers. Each application generates logs locally. How do you find a problem that spans multiple services?

Centralized logging is the answer! It's a system that collects, processes, and stores logs from all your applications in one accessible place.

When logs stay on individual servers, finding issues becomes a nightmare. You'd have to:

• Log into each server separately.
• Search through potentially huge, unstructured log files.
• Manually correlate events across different machines.

This is slow, error-prone, and nearly impossible in modern distributed systems.

A typical centralized logging system has several key components working together. Think of it as a pipeline for your log data.

The main parts are:

• Agents: Collect logs from applications.
• Collectors/Aggregators: Process and enrich logs.
• Storage: Store logs for long-term retention and search.

Agents are lightweight programs installed on each server or within each application container. Their primary job is to watch for new log entries and send them to the next stage.

Popular examples include Filebeat, Fluent Bit, and rsyslog.

Agents do more than just read files. They can:

• Tail log files: Read new lines as they're written.
• Read from standard output/error: Capture console logs.
• Buffer data: Store logs temporarily if the destination is unavailable.
• Add basic metadata: Like hostname or IP address.

They are designed to be efficient and use minimal resources.

After agents, logs often go to a Collector or Aggregator. These are more powerful components designed to receive logs from many agents, process them, and prepare them for storage.

Examples include Logstash, Fluentd, and Vector.

Collectors perform crucial tasks to make your logs useful:

• Parsing: Extracting meaningful fields from unstructured log lines.
• Filtering: Dropping irrelevant logs or specific fields.
• Enrichment: Adding more context, like user IDs or geographic data.
• Routing: Sending logs to different destinations based on their content.

Once processed, logs are sent to a Storage layer. This is where your logs reside for querying, analysis, and long-term retention.

Key characteristics of good log storage:

• Scalability: Handles huge volumes of data.
• Searchability: Allows fast, complex queries.
• Durability: Ensures logs aren't lost.

Popular choices include Elasticsearch, Splunk, and cloud object storage like AWS S3.

Having logs stored is only half the battle. You need tools to explore and visualize them! This usually involves a User Interface (UI) that connects to your storage.

Tools like Kibana (for Elasticsearch) allow you to search, filter, create dashboards, and set up alerts based on your log data.

Let's put it all together. A log entry typically follows this path:

1. An Application generates a log message.
2. A Logging Agent collects the log from the application's host.
3. The agent sends the log to a Log Collector/Aggregator.
4. The collector processes and transforms the log.
5. The collector forwards the processed log to Log Storage.
6. A user uses a Visualization Tool to search and analyze the stored logs.

In this lesson, we explored the architecture of centralized logging systems. You learned about the crucial roles of:

• Logging Agents: Collecting logs efficiently.
• Log Collectors/Aggregators: Processing and enriching logs.
• Log Storage: Providing scalable and searchable log repositories.

This setup allows for efficient troubleshooting and analysis across complex distributed systems. Next, we'll dive into practical methods for collecting and parsing these logs!