LoongCollector is a next-generation data collector for Simple Log Service (SLS) that combines performance, stability, and programmability for building observability pipelines. Unlike traditional log collectors limited to a single scenario, LoongCollector collects, processes, routes, and sends logs, metrics, traces, events, and profiles through a unified agent.
Collection scenarios
LoongCollector supports the following data collection methods:
-
Collect text logs from hosts: Collect logs from files on hosts.
-
Collect container logs from Kubernetes clusters: Collect logs from containers in your clusters.
-
Collect HTTP data: Collect the body content of HTTP requests.
-
Collect SQL query results: Collect the results of specified SQL queries.
All supported scenarios are listed in Log data collection.
Core advantages
On-device collection integration
A single LoongCollector agent handles logs, metrics, traces, events, and profiles end to end. In Kubernetes environments, it uses the Container Runtime Interface (CRI) API to automatically attach metadata labels such as Namespace, Pod, and Container to collected data without changing container configurations.
Flexible programmable pipelines
LoongCollector combines the Structured Process Language (SPL) with multi-language plugin engines into a fully programmable pipeline system.
Choose native plugins for execution efficiency and supplement them with extended plugins, or use the SPL engine for comprehensive operator coverage over complex data.
|
Programmable engine |
Classification |
Features |
|
Native plugin |
C++ implementation. High performance, low resource overhead, and relatively complete operator coverage. |
|
|
Extended plugin |
Go implementation. High performance, low resource overhead, and relatively complete operator coverage. |
|
|
SPL engine |
C++ implementation. High performance, low resource overhead, columnar model with vectorized execution, 100+ operators, and a pipeline design for complex data processing. |
High performance and reliability
LoongCollector delivers stable, efficient observability data collection for large-scale distributed systems.
-
High performance: Lock-free core processes, event-driven architecture, and single-thread throughput of hundreds of MB/s in core scenarios.
-
High reliability: Time-slice fair scheduling, queue-watermark backpressure control, multi-tenant pipeline data isolation, and zero-copy data streaming.
-
Production-grade at scale: The commercial edition is deployed on tens of millions of instances across Alibaba Group, Ant Group, and public cloud customers. Battle-tested during online sales promotions and extreme scenarios.