Databases
Databases shows where application data lives and how Hosting reaches it. Managed WordPress and
Moodle applications receive their database identity through their application setup. An
application built from your code gets one when Add an app or static site
finds it needs a database, or later with Use a database on its
Settings tab; either way the connection settings are delivered to it.
A database declared in hosting.yaml is described in Configuration.

Each application has its own identity
Each managed WordPress or Moodle application gets its own database and its own database user on the environment's shared MariaDB server. Each database identity is restricted to its own database. This does not isolate applications' files: an environment maintainer can read sibling configuration files and their database credentials. See SFTP & SSH access.
For each application the page shows:
- Internal host:port — the address the application uses, reachable from inside the environment. MariaDB is a private service and is never exposed to the internet.
- Database — the database's own name.
- Username — the identity's user name.
- Password — revealed only when you ask for it, with Show password.
Copy a password with the copy control beside it. The page does not keep it visible: hiding it again, or leaving the environment, clears it.
An application's database credentials belong to that environment. A copy you paste into a staging environment is a different database, even when the names look alike. Treat the credentials as secrets: they are enough to read and change everything that application stores.
Temporary-table memory
The environment's MariaDB is shared by every application in it, and one tuning control bounds the memory a query may use for its internal temporary tables — the tables sorting and grouping build while a query runs. Empty means automatic: the engine keeps its own default (16 MiB).
The server decides that limit as the smaller of two engine settings, so this one control writes both. Exceeding it is not an error: the query writes its temporary table to disk instead, which is slower. A larger value is not automatically faster, and a value that is too large for a small database leaves less memory for the data cache.
- An explicit value stays fixed when you change the database's RAM; the automatic memory values are recalculated instead.
- The limit is reserved inside the database's own memory, so raising it reduces what is available for the InnoDB buffer pool.
- This database serves every application in the environment: a change applies to all of them, and open database sessions may be interrupted when the platform applies it.
- While you have a full MariaDB configuration set, that text replaces the generated settings and the guided values are inactive; they stay saved and take effect again when the text is cleared.
Run SQL in the browser
Each application's database panel links to Adminer, signed in as that application's database user. It opens in its own browser tab and reaches only that application's database — the identity it signs in with does not have access to its siblings. Adminer speaks MariaDB, MySQL and PostgreSQL, so every application database has the same entry point.
The cache
Where an environment runs a Redis cache, the endpoints that actually exist are shown beside the databases, each with the sentence that belongs to it. A cache that requires TLS reports only the secure endpoint; one that leaves plaintext open reports both and says what the plaintext address costs. A cache with no credential says so plainly, because any application in the environment can then read and write it.
The cache is not a database: treat anything in it as disposable, because it can be lost and rebuilt at any time — unless you chose the append-only file, which keeps a copy through a restart. It is reachable from inside the environment only. The cache is a single instance that runs inside your environment's shared Pod, so it restarts when that Pod is replaced. See The cache (Valkey) for how an application connects and its limits.
Restoring a database
Databases are backed up — see Backups and restores — but a restore is not a self-service action. A database restore is done by us, from the backup track, into a reviewed target.
You can take your own temporary copy of a database at any time: see Taking a copy of a database.
Retiring a database
When an identity is no longer used, a Company Admin can retire it instead of leaving it accepting logins. Retirement is a reviewed action, not a delete:
- Open the identity and choose Review retirement. The review names the exact database and SQL user, the applications that consume it, and what is deliberately kept.
- Nothing is retired while an application still consumes the identity. Detach every application that consumes this database first: retirement disables NEW logins, and it does not stop or unbind the applications still using it.
- Confirm with the acknowledgement. The confirmation is given against the reviewed identity and the environment's current change generation, so if either changed since you read it the request is refused and you review it again.
Retirement disables new SQL logins only. It does not delete the database or any row, and it does not end SQL sessions that are already open. Its database name, SQL user name and credential version are retained, and its backup track keeps its existing retention.
The identity's Status then reads its real state:
- Retirement pending — the request is recorded and the platform has been asked to apply it. Nothing is retired until the SQL server is observed refusing new logins.
- Retirement failed — the retirement could not be confirmed, so the login state is unconfirmed. The data and existing SQL sessions are unaffected: resolve the reported failure, then request it again through a fresh review.
- Retired — new logins are disabled and observed. The data and backups are retained.
A retiring identity is no longer offered wherever an active database is selected, and its password is no longer revealed. To use the data again, ask for a restore from its backup track.
Next
- Change how much CPU, memory or storage the database may use: Settings and resources.
- Ask for a database restore: Backups and restores.