A host move is a launch with history behind it. The site is already indexed, and the mistakes land on URLs people use.
Before the move
1. record every dns record, including mx and txt
2. confirm the old host can still serve the site for a week
3. list every subdomain, including the ones nobody remembers
4. copy the sitemap and the current status of each url
5. confirm who owns the domain, and who can change dns
During the move
Change one layer at a time: files, then data, then DNS. Lower the time to live a day before, not during.
A worked log
Cedar and Slate Interiors moved from a shared host to a platform. Three things broke.
09:10 files copied, preview host serves the site
09:40 certificate issued for the apex and www
10:05 dns lowered to 300 seconds the previous day, so propagation is quick
10:20 apex resolves to the new host
10:25 break 1: email
mx records still point at the old host, which no longer relays
fix: restore the five mx records from the pre-move copy
time lost: 50 minutes
11:20 break 2: forms
the endpoint allowlists the old host's ip, so submissions are refused
fix: add the new egress addresses, then submit a test response
time lost: 25 minutes
13:00 break 3: redirects
/project-oak was redirected by the old host, and the new host
has no rule for it, so it 404s
fix: add the twenty-seven redirects from the frozen map
time lost: 40 minutes
checks after
- home page, three deep urls: 200
- form test: stored, notification accepted
- test email: delivered, mx records verified
- sitemap urls: 131 fetched, 2 404s, both fixed and re-fetched
- analytics tag: fires once per page
What usually breaks
Email, because MX records are easy to forget. Forms, because an endpoint was allowlisted by address. Redirects, because the old host was doing them.
Acceptance: the move is finished when
- email sends and receives, verified from outside your network
- a test form submission is stored and a notification is accepted
- every URL in the frozen map resolves, in one hop
- one canonical host answers, and the other redirects
- the certificate covers every host you serve
- the sitemap lists only URLs that return 200
Troubleshooting
| Symptom | Likely cause | What to do |
|---|---|---|
| Mail bounces after the move | MX records were not copied | Restore them from the record |
| Forms return an error | The endpoint allowlists the old addresses | Add the new ones, then test |
| Some visitors see the old site | DNS time to live was long | Wait, and watch which host serves them |
| A page 404s that used to work | The old host did redirects | Add the rule to the new host |
| Search impressions dip for two weeks | Normal for a move | Compare 28 days with 28, and check the log |
Limits
A DNS change is not instant, and it is not uniform while it propagates.
Keep the old host answering for at least a week. Watch for traffic arriving on paths you had forgotten existed.