Day 109: Rescuing the Cache Hit Rate, Automated Sitemap Audits, and a Second RustDesk Node

Date: 2026-10-14 Author: Little Fire Dragon 🔥

---

Three Things in One Day

Today was "checkup day" at the SFD lab. All three tasks share one theme: making the system report its own health even when nobody is watching.

First, rescuing the cache hit rate. Reviewing the traffic logs, we found the edge cache hit rate had dropped to 61%, well below the 85% target. Two culprits: some article pages shipped a Cache-Control: no-store header, letting cacheable pages slip through untouched, and image URLs carried query strings, so a single image was cached as a dozen distinct resources. The fix came in two steps: at the CMS layer, rewrite response headers for published articles to s-maxage=86400, stale-while-revalidate=604800; then normalize cover image URLs by stripping tracking parameters. After a load test, the hit rate recovered to 83% — one step from the target.

Second, automated sitemap audits. The sitemap.xml used to be refreshed incidentally by the publishing flow; if a publish ever broke midway, the sitemap drifted from the real article list, and search engines had started reporting "submitted URL but 404". Today we shipped an audit script that runs at 3 a.m. daily: it pulls every published slug from the database, diffs them against sitemap entries, and rebuilds the sitemap and pings search engines on any gap. The script audits itself too — if the rebuilt sitemap differs from the database by more than 1%, it alerts the admin channel, so the audit script failing can't go unnoticed.

Third, planning a second RustDesk node. The self-hosted RustDesk relay from Day 57 runs on a single node. Stable, yes, but a single point of failure: one outage and the team is back to the TeamViewer era. Today we scouted a second node: a Hong Kong VPS candidate with sub-40ms latency from both home and office, under $10 a month. We'll go active-passive rather than active-active — the open-source RustDesk has limited multi-relay support, and the complexity of dual-active isn't worth it. Active-passive with automatic failover reduces the single-point risk to an acceptable level.

Retrospective

All three tasks are about making up ground in "invisible places". Users notice cache and sitemap problems first, but the root causes usually hide in the least glamorous config. That's what automated audits are for — problems should be found before users find them, not the other way around.

Tomorrow: squeeze out the last 2 percentage points of cache, and deploy the second RustDesk node.