Title: 以 OpenClaw 为例，面向 Agentic 应用的统一可观测实践丨 XCOPS 广州站 | BestBlogs.dev
URL Source: https://www.bestblogs.dev/article/6d736dc4
Published Time: 2026-04-12 23:16:00
Markdown Content: 81
This article previews the topic that Liu Haoyang, an observability technology expert from Volcano Engine, will share at XCOPS Guangzhou Station, focusing on solving the observability challenges faced by Agentic applications during large-scale deployment and introducing a unified observability practice solution based on OpenClaw. d dbaplus社群
Yesterday 894 words (about 4 min) View Source →
Sign in to highlight text and take notes as you read. Sign in now
dbaplus社群 2026-04-13 07:16 广东
聚焦智能体应用规模化落地过程中，传统微服务可观测体系无法适配的核心痛点。 2026 XCOPS智能运维管理人年会将于5月22日在广州举办，本次大会聚焦AI时代的真实落地实践，紧扣大模型迭代、Agent深度应用等技术热点，邀请行业领军人物、资深技术专家与学术大师，从技术架构、实战案例到科研成果，与大家一起探索AI应用于智能运维与数据库的最佳方式，找到可落地、可复用的破局方案。其中，火山引擎可观测技术专家刘浩杨老师将分享《以OpenClaw为例，面向Agentic应用的统一可观测实践》，一起来先睹为快：
XCOPS · 广州站 以OpenClaw为例， 面向Agentic应用的统一可观测实践 议题简介：
- 了解Agentic应用与传统微服务的可观测核心差异，精准锚定智能体应用观测的核心发力点；
- 掌握Agentic应用全链路可观测的体系化解决思路，破解大模型调用、工具调用、多智能体协同等新场景下的链路断层、根因定位难、故障复现难的核心痛点；
- 学习大规模生产环境下，Oneagent的MTL统一采集、高性能优化、探针全生命周期管控的实战经验，解决多源观测数据割裂、采集性能损耗过高、探针运维混乱的工程落地难题。
!Image 2: 刘浩杨200x200圆.png 刘浩杨 火山引擎 可观测技术专家 讲师介绍：
* 目前就职于火山引擎，担任可观测性数据采集与集成基础设施技术负责人，主导建设的可观测采集平台覆盖数百万级服务器，单日承载处理数十PB可观测数据，具备超大规模分布式架构的全链路落地经验；
* 在可观测性领域深耕多年，对传统微服务到新一代Agentic智能体应用的可观测体系建设有完整的技术沉淀与实战积累。
以上议题将会在2026 XCOPS智能运维管理人年会-广州站完整呈现，更多互联网大厂及金融代表企业在“垂类Agent应用与人机协作模态”、“数据库自治与底层技术演进”、“金融核心改造与安全效能双升级”等方向上的最新研究与最佳实践，都可以在5月22日的XCOPS广州站一次性看全。 大会议程
!Image 3 码上报名
!Image 4 **码上报名，享早鸟优惠** ↓点这里了解大会更多详情及报名 阅读原文 跳转微信打开
Key Quotes
> Focus on the core pain points where traditional microservices observability systems fail to adapt during the large-scale deployment of agent applications.
> Master the systematic solution for end-to-end observability of Agentic applications, solving core pain points such as link breakage, difficult root cause localization, and difficulty in fault reproduction in new scenarios like LLM calls, tool calls, and multi-agent collaboration.
> Learn practical experience with MTL unified collection, high-performance optimization, and full lifecycle management of Oneagent probes in large-scale production environments, solving engineering challenges such as fragmented multi-source observability data, excessively high collection performance overhead, and chaotic probe operations.
Tags
Agentic Applications
Observability
LLMOps
Intelligent Operations
Volcano Engine
Related Articles
* Production-Grade Nginx Optimization Practices: Core Configuration, Performance Tuning, and Key Issues * Linux 7.0 Released: Google's Million-Dollar Legacy Bug Finally Fixed, Rust Officially Promoted * The Core Framework Behind OpenClaw, Pi: A Good Coding Agent Should Let Users Decide What They Need * Building a Claude Code from Scratch: A Comprehensive Summary! * Tuning a 10-Billion-Scale ES Cluster: Deleting 1B+ Data Points and Reducing Query Latency from 5s to 200ms... * Major Pitfalls in Container Runtime Migration! Netflix's CPU Architecture Bottleneck Investigation and Fixes * #459. Infrastructure Reimagined in the Era of AI Agents: LangChain Founder Unveils the Future of Harness, Memory, and Autonomous Operation * Data Latency Reduced by 8x! Trip.com's Practice in Building a Near-Real-Time Lakehouse from T+1 to Minute-Level * Is Your AI Agent Truly Under Control? * Stop Talking About '10x Developers': AI Agents Aren't Accelerating the SDLC, They're Ending It, reshaping it into a tight closed loop centered on 'Intent, Build, Observe.'")