Skip to main content

Backups and restores

Your environment is backed up on two independent tracks:

  • the application files — the trees each application runs from;
  • the managed databases your applications use, including MariaDB, MySQL and PostgreSQL.

Both are listed on Hosting → Backups → Status, each with its last success and what was observed. A backup runs frequently, but retention keeps one restore point per day: a long retention gives you more days to return to, not every run of the day.

A safe value is one that ran

The page shows what was observed, not only what is scheduled. A track that has never run, or whose last success is not reported, says so rather than showing a blank that reads like success. A blank is not a backup.

An environment's Backups page

Retention​

The retention panel states, in plain numbers:

  • Included in the package — the daily restore points your package includes.
  • Purchased extra days — any extension you have bought.
  • Effective retention — what is actually in force today, and its ceiling.
  • In force since — when the current agreement started.

Changing retention​

A Company Admin can Request a retention change. It is a quote request: we review it, and nothing changes until it is approved. A reduction is confirmed deliberately, with the restore points at risk named before it is submitted, and it applies on the date you choose — older restore points can be removed once that date arrives.

While a cancellation is open, retention changes are refused: the cancellation already promises a date, and changing the retention underneath it would move the goalposts.

Your own bucket as a second copy​

A Company Admin can point your backups at your own S3-compatible bucket as well as ours, from Company → Storage → Backup destination. The steps are in Object storage. This page covers what then arrives there.

  • It is a second copy, never the only one. Our copy stays the one our retention promises cover and the one we restore from. Your bucket is yours to keep, expire or delete from — we never read it back to restore you, and deleting an environment does not reach into it.
  • You receive the same archive we keep, written under the folder you chose and then the environment's name: a restic repository holding the application files, plus the database backup objects your engines write (PostgreSQL's base backups and WAL, and the MySQL/MariaDB dumps). It is a copy of our backup, not a separate export, so it grows with every backup we take.
  • The file archive needs its repository password. It is a restic repository, so restoring it takes restic and the repository password. Ask support@haak.co for it, naming the environment; we hand it over to a verified Company Admin. Keep it somewhere safer than the bucket — a copy you cannot decrypt is not a backup.
  • The database objects need the engine's own tools, the same ones listed under Taking a copy of a database.
  • Copies are written on their own schedule, not instantly: a copying pass runs every fifteen minutes and works through your environments in turn. Each environment's own copy state is shown with its backups.
  • You can ask for one now. On Hosting → Backups → Fresh backup, Ask for a fresh backup takes a new application-file backup and a new managed-database backup of everything that environment holds, then copies the result to your bucket. It captures the platform's existing application-file and managed-database backup tracks and copies their archive; it does not reach into external buckets, remote services or other separately stored data. It starts immediately and finishes when the transfer finishes; the request itself shows queued, in progress, complete or a failure you can act on. You do not need to wait for the fifteen-minute pass, and asking while another request is already running for that environment is refused by name rather than queued twice.
  • A fresh request is a fresh copy, not the last one. Completion is only reported once a copying pass that started after your captures finished has finished, so a copy already in progress when you asked never stands in for the new one.
  • Nothing about the platform's copy changes. Our authoritative copy, its retention and the restores we can run from it are exactly as they were; a fresh request adds a copy in your bucket and replaces no restore point here.
A final copy at deletion

When an environment is deleted and you have a bucket, we copy the archive once more first. If your environment's archive is larger than one copying pass, that last copy stops part-way: the objects already written stay in your bucket, and the deletion report says the copy did not finish rather than claiming it did.

Asking for a restore​

One application is restored from its own page, by a Company Admin: open the application, choose Backups and restore, pick one of the restores it offers, read what it replaces, and confirm. Restore patterns covers the patterns and when each one is right.

Everything else — a whole-environment rollback, or going back to an earlier day than the latest backup — is done by us. Email support@haak.co with three things:

  1. the environment — the page prints the environment's own identifier on Overview → Troubleshooting for exactly this;
  2. which application or database you need back;
  3. the time you want to return to.

We confirm which restore point is still available, then tell you when the restore starts and when it finished. If the restore point you want is older than the effective retention, it is already gone and cannot be recovered by extending retention afterwards.

You can download a retained database backup or file snapshot yourself, or take a fresh database copy — see Downloading a retained backup. Reactivation is not self-service: anything beyond one application's reviewed reset/restore, a database copy you take yourself, and withdrawing a cancellation is arranged with us — see Cancelling hosting.

Downloading a retained backup​

A Company Admin can download a retained database backup or application-file snapshot:

  1. Open Hosting → Backups → Exports and choose Export a retained backup.
  2. Wait for the retained backups to be listed, then choose the capture you need.
  3. Check Includes, Captured, File format and Download expires, then choose Prepare download.
  4. The export appears in Backup exports. When it is Ready, choose Download.

Each choice names the exact recorded database backup or file tree. An environment file export lists its included client trees; SSH host keys and other platform system files are excluded. Database and file backups have separate capture times, so they are not promised to be one simultaneous application snapshot. Files arrive as a .tar.gz archive, which opens with tar; a captured Valkey snapshot remains dump.rdb inside that archive. Database files use the portable formats listed below.

Retained exports normally stay downloadable for two hours from the request, including preparation time. The dialog shows the actual deadline. A capture too close to the end of its allowed retention is refused instead of extending retention or substituting a newer backup. If a capture disappears or expires during preparation, the export fails with its reason and temporary resources are cleaned automatically. A partial source-listing failure is shown beside any database backups that remain available.

From the terminal​

The same operations are available in hostingctl:

hostingctl environment export-sources-refresh <environment-uuid> --company acme --confirm
hostingctl environment export-sources <environment-uuid> --company acme
hostingctl environment export-create <environment-uuid> '<source-id>' --company acme --confirm
hostingctl environment database-exports <environment-uuid> --company acme
hostingctl environment database-export-download <environment-uuid> <export-uuid> --company acme

Refresh starts discovery in the background; the source-list command reports its progress. Use the exact source ID it returns when requesting the export. The export inventory is shared with fresh copies.

Taking a copy of a database​

A Company Admin can take a copy of one database without asking us: open the database on Hosting → Databases and choose Create backup now. The copies that exist are listed on Hosting → Backups → Exports, each with its scope, size, state and download deadline. A Member can see that list; creating a copy, downloading it, or removing it is a Company Admin action.

The file is in the engine's own portable format, so it restores into an empty database of the same engine. Which database engine your environment runs decides both the file and the tool that opens it:

EngineFileOpen it with
MariaDBgzipped SQL dump (.sql.gz)mysql
MySQLgzipped SQL dump (.sql.gz)mysql
PostgreSQLcustom-format dump (.dump)pg_restore
  • The copy is temporary. The page shows when download access ends; the file is removed by the next hourly cleanup. Ask again for a fresh copy when you need one.
  • The download link is short-lived. Each download mints a fresh link for that one file. The link is not a password and not a permanent address, and it is never listed on the page.
  • One copy at a time. A second request while a copy is being made is refused, because two dumps would compete for the same database.
  • You can remove a copy early. Removing one deletes the file; the record of what was handed over stays.

Fresh database copies normally remain downloadable for seven days. Retained exports use their own stated deadline. Download access ends at that deadline; physical files are removed by the next hourly cleanup. You may keep a copy you have already downloaded.

A request is refused, with the reason stated, when a copy is already being made, when the copy would outlive the retention you were told, when the database identity is retired, or when the engine cannot be exported.

Restoring into a new environment​

If you need the data without disturbing the running environment, say so in the request: we can restore into an isolated environment and hand you the result. This is the same staff path as any other restore, and it is the one to ask for when you are recovering from a mistake rather than rolling back production.

Next​