SRE on BvBeJ的小站https://www.bvbej.com/tags/sre/Recent content in SRE on BvBeJ的小站Hugozh-CNSun, 24 May 2026 00:00:00 +0000Kubernetes etcd 碎片整理与维护窗口设计https://www.bvbej.com/posts/k8s-etcd-defrag-maintenance-window/Sun, 24 May 2026 00:00:00 +0000https://www.bvbej.com/posts/k8s-etcd-defrag-maintenance-window/控制面稳定性依赖 etcd 健康，碎片整理要纳入常规运维节奏。Kubernetes Priority 与抢占：如何守住核心服务 SLOhttps://www.bvbej.com/posts/k8s-priority-preemption-slo/Wed, 20 May 2026 00:00:00 +0000https://www.bvbej.com/posts/k8s-priority-preemption-slo/优先级体系设计不好，抢占机制会从保命手段变成故障放大器。Service Mesh 的超时与重试预算：避免放大故障https://www.bvbej.com/posts/service-mesh-timeout-retry-budget/Sat, 09 May 2026 00:00:00 +0000https://www.bvbej.com/posts/service-mesh-timeout-retry-budget/重试不是越多越稳，预算化治理才能防止雪崩扩散。Kubernetes 中 HPA 与 VPA 如何协同而不互相打架https://www.bvbej.com/posts/kubernetes-hpa-vpa-coordination/Thu, 30 Apr 2026 00:00:00 +0000https://www.bvbej.com/posts/kubernetes-hpa-vpa-coordination/水平扩缩容与资源建议同时启用时，如何避免抖动与反复重建。OpenTelemetry 指标治理：控制基数比盲目埋点更重要https://www.bvbej.com/posts/opentelemetry-metrics-cardinality-control/Sat, 25 Apr 2026 00:00:00 +0000https://www.bvbej.com/posts/opentelemetry-metrics-cardinality-control/高基数标签会直接拖垮指标系统，如何从设计上避免