支柱：基础的可观测性要素 Metrics, tracing, and logging 2017.02.21 Peter Bourgon simplify the growing complexity © 2021, YUNSHAN Networks Technology Co., Ltd. All rights reserved. 开箱即用的解决方案 Metrics, tracing, and logging 2021, YUNSHAN Networks Technology Co., Ltd. All rights reserved. 2.0 服务：统一的可观测性平台 可观测性平台（Metrics、Tracing、Logging） 基础设施团队 业务团队A 业务团队B 业务团队C 业务团队D …… 存储、检索服务 观测数据 观测数据 观测数据 观测数据 simplify the growing complexity 300x 100K 300K 100M DeepFlow的演进之路 (每秒列读写速率) InfluxDB+ DeepFlow 5.5 30x 2M 18mon 18mon 18mon Metrics Tracing Logging Tag, Exemplars Tag Tag, Exemplars Tag Tag, TraceID Tag, TraceID “Index-free” “Index-free”

SkyWalking 历史和特 点 2/28SkyWalking 简介 /01 SkyWaling 的历史和特点 3/28Micro Service 4/28End to End Distributed Tracing 5/28Apache Way 6/287/总页数Architecture 8/28分享主题：字号 分享嘉宾 服务 抽象概念，用于汇集指标 Service 实例 进程，容器，Pod 11/28Istio 1.5 架构图 12/28挑战1：技术路线多变 基于 Log 成熟、但性能低 Mixer 基于 Metric 高效、但技术门槛高 Mixerless 13/28挑战2：无 Tracing VS 14/28挑战3：维度匹配-Mixer Instance Service Endpoint 15/28挑战3：维度匹配-Telemetry2 Instance Service AccessLogService 22/2823/总页数 技术路线全覆盖-TelemetryV2(WIP) 23/28Tracing-协议支持 24/2825/28 Tracing SkyWalking 8.0之前 Tracing，Metrics 只能二选一 SkyWalking8.0 Tracing，Metrics 可以同时支持维度匹配 SkyWalking Mixer Mixerless Service

Hybrid profiling capabilities (sampling and/or instrumentation) (CPU and GPU instrumentation) Tracing capabilities (values, messages, plots, allocations, …) Hassle-free integration (single source* macOS, iOS, Android, WASM*) Hybrid profiling capabilities (sampling and/or instrumentation) Tracing capabilities (values, messages, plots, allocations, …) Free and Open source Kudos to Bartosz Taudul macOS, iOS, Android, WASM*) Hybrid profiling capabilities (sampling and/or instrumentation) Tracing capabilities (values, messages, plots, allocations, …) Free and Open source Kudos to David Farrell

Hybrid profiling capabilities (sampling and/or instrumentation) (CPU and GPU instrumentation) Tracing capabilities (values, messages, plots, allocations, …) Hassle-free integration (single source* macOS, iOS, Android, WASM*) Hybrid profiling capabilities (sampling and/or instrumentation) Tracing capabilities (values, messages, plots, allocations, …) Free and Open source Kudos to Bartosz Taudul macOS, iOS, Android, WASM*) Hybrid profiling capabilities (sampling and/or instrumentation) Tracing capabilities (values, messages, plots, allocations, …) Free and Open source Kudos to David Farrell

功能/非功能仍需加强Contour & Istio • 南北向流量 • API版本共存（Istio & Kubernetes Ingress） • 东西向流量 • Istio支持全量功能 • Contour支持Tracing能力 • 数据面共存 • 共用边车组件Envoy七牛现有Service Mesh体系 • Istio产品化 • 东西流量产品化 • 南北流量产品化 • TLS管理优化 • Contour增强 • Istio的灰度发布和流量管理 • Istio的Tracing产品化落地场景—云存储系统 • 历史问题 • 灰度发布 • 预上线系统验证体系 • 系统故障隔离 • 跨集群访问 • 线上问题链路追踪 • 解决方案 • Istio南北流量分流策略产品化 • 基于Istio的QoS产品化 • 跨集群流量调度 • 基于Istio的Tracing产品化落地场景—大数据产品 • 系统优化之路 • • 多版本灰度升级 • 根据流量做横向伸缩 • 分布式系统性能测试问题 定位难 • 解决方案 • Istio南北流量分流策略产品化 • 基于Istio的QoS产品化 • 基于Istio的Tracing产品化 • 跨集群流量调度七牛容器云产品逻辑架构 Kubernetes + Docker + 生态链七牛容器云产品未来发展 • 更多功能 • 故障熔断 • 故障注入 • 业务配置动态分发 •

backwards • Init…uninit • Load…unload • Start…shapshot-save…shutdown • Lifecycle Exception or Error Tracing • Source…sink • Producer…consumer • Push…pull (how to leak back failed operation from previous backwards • Init…uninit • Load…unload • Start…shapshot-save…shutdown • Lifecycle Exception or Error Tracing • Source…sink • Producer…consumer • Push…pull (how to leak back failed operation from previous backwards • Init…uninit • Load…unload • Start…shapshot-save…shutdown • Lifecycle Exception or Error Tracing • Source…sink • Producer…consumer • Push…pull (how to leak back failed operation from previous

Bitmain tech expert Service Mesh Meetup #4 上海海站 2018.11.25Observability CNCF LandscapeMetric, Tracing, LoggingOld game to observeService Mesh is comingProxy and SidecarIstio + Envoy Representative Service out of process adaptor Bypass adpator Adaptor In process Bypass adaptor SkyWalking backend Tracing Metric Receiver in gRPC/HTTP Analysis Core Query CoreIstio telemetry Attribute Vocabulary https://istio

Trace Viewer - https://ui.perfetto.dev/ - Can build and run server locally as well - Chrome Event Tracing Format JSON - … and more!20 VisualizationClang -ftime-trace - Generates JSON file based on output Deeper 22Visualization 23Visualization 24ninjatracing - Convert .ninja_log files to chrome tracing format - Embed -ftime-trace .json files 25Combining Visualizations 26Single File 27 Project

OSP client Pod OSP client Web Server Reg agent Proxy address File 服务注册 宿主机 物理机 物理机 Tracing system Trace-log Trace-log 用户与Istio的区别 我是作者名称没有对称的server端agent • 性能考虑 • Proxy已经增加了一跳，server再增加一跳会加剧 Proxy问题，可立马切走流量而不影响业务代码埋点 vs. Mixer • 性能考虑 • 调用链埋点的影响必须足够小 • 鉴权需要同步进行，调用Mixer代 价大 • Mixer的中央节点问题 • 传统基于日志收集的tracing方 案足够成熟 • 内部实现一套可插拔的鉴权框 架也能接受混合部署 vs. 绑定K8s • 历史原因导致长期都会物理机 和容器并存，内部需求必须要 同时支持物理机和云 • 绑定K8s能够享受K8s的红利，

[--teleporterpassword ] [--teleporterpasswordfile |stdin] [--tracing-enabled on|off] [--tracing-config ] [--tracing-allow-vm-access on|off] [--usbcardreader on|off] [--autostart-enabled dumping it to a file, for debugging purposes. With --nictrace<1-N> on|off, you can enable network tracing for a particular virtual network card. If enabled, you must specify with --nictracefile<1-N> these settings. • --tracing-enabled on|off: Enable the tracebuffer. This consumes some memory for the tracebuffer and adds extra overhead. • --tracing-config : Enables tracing configuration.