Infra Observability Overview
SelectDB Infra Observability is based on the DOG Stack architecture, providing an open, high-performance, and cost-effective observability solution that supports unified collection, storage, and visualization of logs, traces, and metrics.
What is Infra Observability
Infra Observability is the infrastructure observability solution provided by SelectDB, built on the DOG Stack architecture to help you comprehensively monitor and analyze the running status, performance metrics, and anomalies of SelectDB instances.
DOG Stack consists of the following three core components:
D (SelectDB): Responsible for efficient data storage and fast query, serving as the storage and compute engine for observability data.
O (OpenTelemetry): Responsible for data collection, writing telemetry data to SelectDB through OpenTelemetry Collector and Doris Exporter.
G (Grafana): Responsible for data visualization, providing log search, trace analysis, and dashboard functions through Grafana and Doris App plugin.
Core Features
Data Collection
OpenTelemetry SDK: Supports multiple programming languages, providing code-free or simple application instrumentation (mainly for Trace).
OpenTelemetry Collector: Collects local files, host metrics (CPU/memory), container/Pod metrics.
Third-party Integration: Compatible with Vector, Fluentbit, Node Exporter, Telegraf, Logstash, Filebeat and other tools, sending data through OTLP and other protocols.
Log Search
Supports both Lucene-style search syntax and native SQL syntax.
One-click correlation analysis between Log and Trace.
Supports table or JSON format for viewing details.
Can view context data before and after specific logs or traces.
Trace Analysis
Filter by service name, operation, tags, and duration range.
Timeline, duration, and span count visualization; waterfall diagram for specific traces.
Sort by timestamp, duration, or span count; fine-grained span filtering.
Dashboard
Supports 20+ chart types including line charts, bar charts, pie charts, heatmaps, and geographic maps.
Supports visual builder or raw SQL code.
Leverages SelectDB multi-table join capability to correlate logs, traces, and metrics in the same dashboard.
Alerting
Supports rule types including no data/error, threshold, and composite threshold; covers metrics such as count, sum, and ratio.
Supports 20+ notification channels including Email, PagerDuty, Slack, Discord, and Webhook.
Supports customizing, formatting, and reusing alert notification message templates.
Note: To use Grafana alerting features, you need to configure public network SNAT for the VPC where the SelectDB instance is located. For configuration details, see Internet NAT gateway.
System Architecture
The Infra Observability system consists of the following four main components:
Data Collection Tools: Collect logs, traces, and metrics. Includes OpenTelemetry SDK (application/AI agent instrumentation), OpenTelemetry Collector Receivers (filelog, hostmetrics, podmetrics), and third-party tools (Vector, Fluentbit, etc.).
OpenTelemetry Collector (with Doris Exporter): Receives telemetry data, processes it (filtering, transformation, enrichment, queuing, batching), and writes to SelectDB through Doris Exporter.
SelectDB (Storage and Compute Engine): Efficiently stores indexed columnar data, providing fast search and analytical queries. Key technologies include inverted index, bloom filter index, zone map index, and native high performance in aggregation, sorting, and join operations.
Grafana (with Doris App Plugin): Provides visualization of logs, traces, and metrics. Plugin enhanced features include Discover (providing Kibana-like search experience) and Trace (providing native trace exploration capability).
Design Goals
Openness: All components are open source, avoiding vendor lock-in.
High Performance: Write throughput can reach hundreds of GB/s; search/query response time for billions of records is sub-second.
Cost Effectiveness: 50%-80% cost reduction compared to ELK or Datadog.
Scalability: Easily handles PB or even EB-scale data volumes.