China Managed Hosting: Rebuild and Zero-Downtime Migration
TL;DR
Two China websites of an industry-leading manufacturer went down in a May server breach — and the recovery showed the survival was pure luck: 3 days of panel license left, zero snapshots, zero account balance. We rebuilt the hosting and migrated both sites with zero interruption; they now run as a fully managed service with daily backups, real restore drills, and 42-metric automated monitoring, at about 60% lower resource cost.
Background
The client is an industry-leading manufacturer whose products run on landmark projects worldwide — and its corporate site doubles as the landing page for live ad campaigns, so downtime is lost money by the minute. The corporate site and a 289-post WordPress site had been hosted since 2021 on a single server under an overseas cloud account, maintained remotely by the US team across time zones.
In May 2026 the server was breached and knocked offline; a third party restored the data. The inventory that followed was the scarier part: the panel license had 3 days left (expiring May 24, 2026), the server had no snapshots at all, and the cloud account balance was zero — one bill away from a second outage, with nothing to restore anything from. The setup could not support long-term operation either: no CDN, no media cloud storage, and comparable resources cost about 60% more billed through an overseas account.
Solution
The client decided to stop patching and hand its China assets to one locally accountable team. The plan went straight from inventory to design: the old box was an 8-core/32GB machine running nearly idle with its disk 90% empty, so the new one was sized at 4-core/16GB — the 60% saving was calculated, not negotiated by cutting specs. Storage was designed the same way: media and backups live in separate buckets, none open to the public; each bucket has its own independent key, so a leak stays contained; and the storage layout mirrors the site's existing media folders, so not one image URL in the articles had to change.
The move itself was a container move, not a rebuild: both sites' files, databases, and account configurations were bundled as one and compared file by file across both ends, with all 2,862 media files verified after sync. The client held the switch rhythm: production and preview ran side by side, the corporate site went through a page-by-page review on its own preview domain, and live traffic moved only after sign-off — with the old server kept on standby, one DNS record away from an instant rollback. Across the DNS switch and both cutovers, visitors never noticed a thing.
Every gap the May intrusion exposed was closed one by one: dual allowlists at the cloud and server layers, remote management reachable only from fixed networks belonging to the client and us; automatic banning of repeated login failures plus brute-force protection on the admin panel; two-step verification enforced on top-level access; and an ops account that, even if stolen, can run exactly one pre-approved command. Team members who don't live in a terminal got a Windows/macOS desktop app — one click for a one-time link into the admin panel, never touching top-level credentials, and the app's own key can do only that one thing.
The filing change required for long-term operation in China was submitted by the client's own entity with our technical materials in support; the production cutover was scheduled after the approval landed.
Media moved to the cloud through our in-house WordPress plugin: originals and every generated size sync into cloud storage, page images point to the CDN automatically, and any failure falls back to the site's local files — the site never loses images because of the cloud. HTTPS certificates for the CDN and the buckets renew themselves; the long-self-signed certificates we inherited were traced to two AutoSSL misconfigurations.
And the luck-driven cycle of May was replaced with a system: our in-house Server Sentinel now keeps watch — 42 health and security metrics checked automatically, key ones every 5 minutes; brute-force attempts, suspicious logins, persistence backdoors, and cryptomining processes inspected daily; backups verified every day and validated by a first real restore drill on September 16 — all 76 database tables restored and compared, every one a match. The old server is still on standby, demoted from "life raft" to "last resort".
Results
| Dimension | Before handover | After handover |
|---|---|---|
| Hosting | Single server, overseas account; license expiring, zero snapshots | Hosted in China, clear resource ownership, ~60% lower cost at equal resources |
| Site trust | Self-signed certificates, browser warnings | Certificates issued and renewed automatically, full CDN coverage |
| Account security | Breached before; single-password protection | Five access layers, admin entrances off the public internet |
| Data assets | No snapshots, no off-site backups | Daily off-site backups; restore drill passed, all 76 tables match |
| Observability | Found out when it broke | 42 metrics auto-checked; monthly patching ~3 minutes, zero incidents |
Efficiency and stability: across the DNS switch and both cutovers, neither site went down; three server-side optimizations (transfer protocol, code caching, compression) brought the homepage from 172 KB down to 33 KB — about 5× smaller. The engagement left behind a migration inventory, bilingual migration plans, seven numbered work logs, and an issues log — all in the client's hands, ready for any third-party review.
Why Us
You hand your China environment to one locally accountable team: the monitoring, the cloud-offload plugin, and the login tool are all built in-house — no black boxes. Numbered logs and bilingual monthly reports mean you don't have to take our word for anything; any third party can audit the work.