Dock

Dock is Slipway's database and cache workbench. It connects to a running service attached to the selected environment and uses that service's real credentials inside the private Docker network.

Dock is not an ORM abstraction. SQL, MongoDB expressions, Redis commands, imports, and generated migrations execute against the selected live service.

Supported services

Service	Console	Browse	Schema	Model diff	Import and export
PostgreSQL	SQL	tables and rows	columns and indexes	full Waterline schema diff	SQL text and PostgreSQL custom dump
MySQL	SQL	tables and rows	columns and indexes	full Waterline schema diff	SQL text
MongoDB	mongosh expression	collections and documents	collection metadata	missing collections	JSON or compressed archive
Redis	redis-cli command	—	—	—	—

Dock appears when the environment has at least one running service of these types. If several services are running, the first screen is a service picker and the selected service ID becomes part of the Dock URL.

Open Dock

Choose Dock from an environment or from a database service action. The production picker uses:

/projects/:projectSlug/dock

Other environments include the environment slug:

/projects/:projectSlug/environments/:environmentSlug/dock

After selecting a service, Slipway appends its ID. Every server request rechecks that the service belongs to the environment, belongs to the signed-in user's team, and is still running.

Console

PostgreSQL and MySQL

Write SQL and use Run or Cmd/Ctrl Enter. Dock executes the query inside the service container with psql or mysql; it does not open the database port to the public internet.

SELECT id, email, created_at
FROM users
ORDER BY created_at DESC
LIMIT 50;

Multi-statement SQL is split and presented as one ordered result per statement. Each result identifies the command, status, duration, row count, affected-row summary, rows, or error. When a later statement fails outside an explicit transaction, earlier successful statements remain visible and may already be committed.

BEGIN;
ALTER TABLE orders ADD COLUMN discount DECIMAL(10, 2);
ALTER TABLE orders ADD COLUMN discount_code VARCHAR(50);
COMMIT;

The server gives console execution 30 seconds and bounds captured process output to 10 MB. Query history is kept only in the current browser page and is capped at 20 deduplicated entries.

MongoDB

MongoDB console input is a JavaScript expression evaluated by mongosh:

db.users.find({ emailVerified: true }).limit(10).toArray()

Dock serializes the returned value into the same table or JSON result surface. Use collection operations deliberately: this is a live shell, not a read-only query builder.

Redis

The Redis console sends one command to redis-cli and displays its raw output, error, and duration:

INFO memory

Quick actions are provided for common inspection commands such as PING, INFO server, DBSIZE, and CONFIG GET maxmemory. Use the up and down arrow keys to navigate the current-page history. The browser keeps at most 200 entries and trims older output after that bound.

Consoles can mutate data

Dock permits ordinary writes. It blocks only a narrow class of catastrophic commands: dropping a PostgreSQL/MySQL database or schema, selected system-table truncation, and MongoDB database/collection drop or shutdown expressions. This is not a general read-only policy. UPDATE, DELETE, FLUSHALL, and many other destructive operations can execute. Back up first and prefer a scoped, reviewable statement.

Read results

Rowsets open as tables and can be viewed as JSON. A multi-statement query adds keyboard-accessible result tabs. Dock keeps command summaries separate from rowsets so an ALTER TABLE or UPDATE result does not look like an empty SELECT.

Copy and export act on the selected result. CSV output quotes commas, quotes, and newlines. Returned values are shown as data and do not execute as page HTML.

Browse tables and collections

The Tables or Collections tab lists objects with an estimated or exact row/document count. Select one to load a page of data.

The browse endpoint defaults to 50 rows, accepts 1–1,000, and supports offset, order field, and ascending or descending order. SQL table and order identifiers must be simple identifiers; Dock rejects names that could inject SQL. MongoDB uses _id as its default order field, while SQL uses id.

Counts can be expensive on very large collections or tables. Use the console with a purpose-built indexed query when the generic browser is not the right tool.

Inspect schema

PostgreSQL and MySQL schema views show the physical table definition, including columns, database types, nullability, defaults, primary keys, and indexes. The table filter is reflected in ?schemaTables= so a focused schema view can be bookmarked.

MongoDB is schemaless. Dock can list collections and compare whether the collections expected by Waterline exist, but it does not infer a field schema from sample documents. Redis has no schema tab.

Compare Waterline models

The Migrate tab compares the selected service with the application's model metadata.

1. When the app is running, Dock first introspects its loaded Waterline models.
2. If runtime introspection is unavailable, Dock falls back to statically reading model source from the pushed build context.
3. Dock reads the service's current schema.
4. It maps Waterline attributes to the physical types used by the selected adapter.
5. It presents grouped, selectable statements for review.

The response says whether model metadata came from runtime or static. Runtime is preferable because it includes the application's effective loaded model definitions. Static fallback allows first-time schema creation when the app cannot lift against an empty database.

What migrations generate

For PostgreSQL and MySQL, Dock can generate:

• table creation;
• column rename when an attribute name maps to a different column name;
• missing column creation;
• compatible column type or nullability changes; and
• missing indexes.

For MongoDB, it creates missing collections only. The generated diff does not drop extra tables, collections, or columns automatically.

An explicit Waterline columnType is used as the desired physical type. Review it carefully because it bypasses the normal logical type mapping.

Apply a migration

Select the model groups, read every generated statement, choose Apply Migration, and confirm the count. Dock executes statements sequentially and stops on the first failure. The response reports each successful statement and the failing one.

Dock does not automatically wrap the whole generated list in a transaction. Earlier statements can remain applied if a later statement fails. Recompute the diff before retrying.

Back up and stage schema changes

Generated SQL is a useful translation of model metadata, not a proof that the data can satisfy a new constraint. Test the same change in staging, inspect existing nulls and duplicates, and create a backup before production.

Initialize an empty database

An application can fail to lift when its production datastore has no tables. Dock can still build a diff from pushed source:

1. push or deploy the application source;
2. create and start the database service;
3. open Dock → Migrate for that service;
4. verify that model source was loaded and review the generated tables;
5. apply the migration; and
6. redeploy or restart the application.

If Dock reports modelsSourceNotFound, push source before trying again.

Export

The export action can include the full database, schema only, data only, or a selected set of tables.

Service	Implementation	Notes
PostgreSQL	pg_dump --no-owner --no-acl	supports table selection, schema-only, and data-only
MySQL	mysqldump --single-transaction --routines --triggers	supports table selection, schema-only, and data-only
MongoDB	mongoexport for one selected collection, otherwise compressed mongodump archive	schema-only does not apply

The current export process has a five-minute timeout and a 100 MB captured output buffer. Large production backups should use Slipway's backup workflow instead of a browser download.

Import

Paste text or upload a supported file, review the destructive confirmation, and then import into the selected service.

Service	Text	Binary/archive
PostgreSQL	.sql through psql with ON_ERROR_STOP=1	.dmp custom dump through pg_restore --clean --if-exists
MySQL	.sql through mysql	not supported
MongoDB	JSON array, optionally with // collection: name	gzip archive through mongorestore --drop

Uploaded PostgreSQL custom dumps must begin with the PGDMP signature. MongoDB archives must be gzip. Slipway rejects a mismatched file before calling the native restore client.

Imports stream through a temporary file or process input so Slipway does not buffer a 500 MB upload in application memory. Defaults are:

Limit	Default
Maximum import	500 MB
Import timeout	30 minutes
Captured stdout	64 KB
Captured stderr	64 KB

The temporary upload is deleted after success or failure. Streaming controls memory use, but the host still needs enough free disk for the temporary file and database growth.

Authenticated endpoints

The Dock UI uses the following session-authenticated routes. Add /environments/:environmentSlug after the project slug for a non-production environment, and pass ?service=:serviceId when selecting a particular service.

Method	Path suffix	Purpose
POST	/dock/sql	execute SQL or a MongoDB expression
GET	/dock/tables	list tables or collections
GET	/dock/tables/:table/data	browse a page of data
GET	/dock/schema	read physical schema
GET	/dock/models	inspect application models
GET	/dock/diff	generate the model/schema diff
POST	/dock/migrate	apply selected statements
POST	/dock/export	export data or schema
POST	/dock/import	import text or an uploaded dump

The common prefix is /api/v1/projects/:projectSlug. These endpoints use the same team and environment authorization as the page; they are not public database APIs.

Troubleshooting

No services appear

Confirm the service is PostgreSQL, MySQL, MongoDB, or Redis; belongs to the selected environment; and has running status. A stopped service is removed from the picker.

A query times out

Console SQL and MongoDB execution has a 30-second timeout. Add a limit, inspect the query plan, add an index, or use a dedicated reporting path for long work.

The diff is empty or wrong

Check the reported model source. Restart the app after model changes so runtime metadata is current, or push the new source so static fallback can read it. Then verify tableName, columnName, columnType, and datastore assignment in the model.

An import is rejected

Check the service, file signature, 500 MB default bound, and available host disk. A PostgreSQL plain SQL file belongs in text mode; .dmp expects the custom pg_dump format.

A migration partially applies

Do not resubmit the old statement list blindly. Reload the diff against the current database, inspect which changes remain, and restore from backup if the partial state is unsafe.

What's next?

• Use Backups before destructive database work.
• Use Helm when the repair is clearer through application models and helpers.
• Use Lookout to correlate database work with application latency and errors.