<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>蓝色札记 · LANUNION ARCHIVE</title><description>网络组档案库 — 运维记录、故障排查与基础设施建设笔记</description><link>https://archive.lanunion.example/</link><language>zh-CN</language><atom:link href="https://archive.lanunion.example/rss.xml" rel="self" type="application/rss+xml"/><item><title>RAID 写惩罚与 IOPS 预算：一块盘掉线之后数据库为何先垮</title><link>https://archive.lanunion.example/archives/raid-write-penalty/</link><guid isPermaLink="true">https://archive.lanunion.example/archives/raid-write-penalty/</guid><description>硬件 RAID5 阵列掉了一块盘，重建刚开始数据库就雪崩。问题不在重建本身，而在这台机器上线时就没有算过小随机写的代价——把写惩罚、可用 IOPS 与重建窗口三个数摆在一起，这台机器的结局在第一天就定了。</description><pubDate>Wed, 16 Sep 2026 00:00:00 GMT</pubDate><category>raid</category><category>iops</category><category>存储</category><category>数据库</category><category>容量规划</category></item><item><title>跨机房备份只跑到 12 MB/s：一次 TCP 接收窗口的排查</title><link>https://archive.lanunion.example/archives/tcp-window-bdp/</link><guid isPermaLink="true">https://archive.lanunion.example/archives/tcp-window-bdp/</guid><description>一条 2 Gbps 的专线，rsync 单连接却钉死在 12 MB/s。用带宽时延积算出这条路径上该有多大窗口，再回头查内核的接收缓冲上限——顺带说明为什么&quot;把窗口开大&quot;在 40 ms 的 RTT 下比在局域网里重要一个数量级。</description><pubDate>Tue, 15 Sep 2026 00:00:00 GMT</pubDate><category>tcp</category><category>带宽时延积</category><category>rsync</category><category>跨机房</category><category>内核调优</category></item><item><title>Docker 自定义网桥间歇性丢包：一次 MTU 收敛排查记录</title><link>https://archive.lanunion.example/archives/docker-network/</link><guid isPermaLink="true">https://archive.lanunion.example/archives/docker-network/</guid><description>同网桥内容器短连接一切正常，大文件传输却在 40 秒后挂起。从 tcpdump 缺失的分片出发，定位到 WireGuard 隧道 MTU 未向容器继承，并给出守护进程、compose 与 MSS 钳制三层的收敛方案。</description><pubDate>Mon, 14 Sep 2026 00:00:00 GMT</pubDate><category>docker</category><category>网络</category><category>MTU</category><category>WireGuard</category></item><item><title>Ubuntu Server 例行巡检笔记：journal 占满根分区与内核待重启</title><link>https://archive.lanunion.example/archives/ubuntu-server/</link><guid isPermaLink="true">https://archive.lanunion.example/archives/ubuntu-server/</guid><description>一份 24.04 LTS 节点的季度巡检记录：快照后清理 systemd-journald 的失控日志、处理 unattended-upgrades 装完内核却没重启的悬挂状态，并用 needrestart 摸清待重启的影响面（其自动重启模式绝不能在数据库节点上开启）。</description><pubDate>Thu, 10 Sep 2026 00:00:00 GMT</pubDate><category>ubuntu</category><category>运维</category><category>systemd</category><category>巡检</category></item><item><title>内网镜像站反向代理整理：缓存、Range 与大文件回源</title><link>https://archive.lanunion.example/archives/mirror-proxy/</link><guid isPermaLink="true">https://archive.lanunion.example/archives/mirror-proxy/</guid><description>自建镜像站上线半年后的配置复盘：用 Nginx 的 slice 模块把大包回源切成固定块、让 apt/yum 的 Range 请求不再重复拉全量，并用 proxy_cache_lock 压掉并发回源的雪崩。</description><pubDate>Tue, 08 Sep 2026 00:00:00 GMT</pubDate><category>nginx</category><category>镜像站</category><category>缓存</category><category>反向代理</category></item></channel></rss>