WordPress backups and restores
How update runs back sites up, when a backup is restored automatically, the emergency FTP restore, and repairing databases damaged by older restores.
Every WordPress update run in OpsDash is backed up, so a bad update can be undone.
How backups work
Choose the backup on the site's settings (Database backup before updates):
- Full (recommended): the whole database is backed up once per run, a little at a time so hosts with short time limits can finish it. Then each update keeps its own files and settings.
- Files only: no database backup, for hosts that keep their own. Each update still keeps its own files and settings.
The run shows the backup's progress ("Backing up the database before the first update…", "Uploading the database backup") and, when done, its size and how long it took.
Automatic restores
After each package, the site is checked from inside and from outside. If the update broke it:
- The site restores its backup straight away, and the step is marked Rolled back.
- If the site couldn't, OpsDash has the site restore its backup.
By default the run then carries on with the next package, as long as the site is healthy again. Tick Stop the run when an update fails on the site to skip the rest instead.
Backups restore exactly, keeping every character (plugin 1.3.4 and later).
Emergency restore over FTP
If an update takes a site down so badly that the connector stops responding and the public site is failing, OpsDash can queue an emergency restore. A separate restore worker (a script, not an AI model) logs in over FTP and puts the backup back.
For this you need, on the site's settings:
- an Emergency FTP login (from the vault) and the FTP root (the path from the FTP login to the WordPress folder),
- an Emergency database login, with the database name on the first line of the login's notes,
and a WordPress restore connection in Settings › Agents, used by the restore worker. It's the only connection that can lease the FTP password, and only while restoring. Agents never receive these passwords.
If updates are held on the site, OpsDash doesn't queue a restore; it tells the site's watchers it needs checking by hand.
Database damage from older restores
Restores made with plugin versions before 1.3.4 could damage some database values (text with a % sign). OpsDash checks each site:
- Restores so far lists restores per site, and what the database looks like now: No damage found, Checking… or damaged values.
- For damaged values, Review and repair it on the WordPress page. Back up, then repair backs the site up first. Updates are held while a repair runs, and carry on once it finishes.
- Check again re-checks at the site's next check-in.
Sites still on an older plugin with this bug have updates held until the plugin is updated, because restoring one of its backups would damage the database.
Run history
Deleting runs from a site's page doesn't delete their backups or restore history.
Settings in OpsDash
These open in OpsDash. You’ll be asked to sign in first if you aren’t already.
Related articles
- What the WordPress connector doesCheck-ins and wake-ups, what the plugin reports, how it updates sites one at a time, the feedback widget, one-click login, MailPoet and self-updates.
- What the update run messages meanPlain-English meanings of the statuses and messages you see on a WordPress update run, and what to do about each.
- WordPress updatesUpdate plugins, themes and WordPress safely, one at a time with a backup, automatically in a maintenance window or by hand, with staging, visual checks and failure tickets.