Package Exports
- @franciscokloganb/appwrite-database-migration-tool
- @franciscokloganb/appwrite-database-migration-tool/cli
- @franciscokloganb/appwrite-database-migration-tool/lib
Readme
Appwrite Database Migration Tool
Disclaimer: This NPM package is in a very early stage, please wait for v1.0.0 release to use on your own projects, unless you want to provide feedback or contribute.
Roadmap to v1.0.0
- Run codegen implementation from a consuming project;
- Run the setup implementation against live Appwrite databases from a consuming project;
- Run the run migration sequence implementation against live Appwrite databases from a consuming project;
- Create one down sequence flow
- Run the one down sequence flow against live Appwrite databases from consuming proeject;
- Add integration tests with mocked Appwrite responses using MSW;
Setting Up
Common/Shared Steps
These steps only need to be made once*.
- Create a
GitHub Repositoryto host your Appwrite functions.- You can use a repository you already own (e.g.,
myproject-functions).
- You can use a repository you already own (e.g.,
- Create
environmentbranches.- For example, the
mainbranch can be assigned toproductionandstagingbranch can be assigned tostaging. This allows you to have multiple Appwrite projects, using a single functions repository containing with multiple serverless entrypoints. Allowing you to effectively test a function in the staging project, before deploying the changes to the main (production) project.
- For example, the
Environment Specific Steps
Functions
These steps need to be done per project that represents an applicatio environment in which you want to use the Appwrite Database Migration Tool processes.
Associate the repository with Appwrite Serverless function source.
Create a Appwrite Database
- Not needed if you already have a Database.
Create an Appwrite Function, one called
Create Migrations Collectionwith the body below.- The function should point at the branch that contains the source for the "environment".
- Tweak the timeout (default is 15s, increasae it up to a maximum of 15m) in the function settings.
import { createMigrationCollection } from '@franciscokloganb/appwrite-database-migration-tool' export default async (ctx) => { await createMigrationCollection({ log: ctx.log, error: ctx.error, }) }
Create another Appwrite Function, called
Run Migration Sequencewith the body below.The function should point at the branch that contains the source for the "environment".
Ensure the migration files you create in the future are included in the final function bundle.
- An example on what this means is given on FAQ section.
import { runMigrationSequence } from '@franciscokloganb/appwrite-database-migration-tool' export default async (ctx) => { await runMigrationSequence({ log: ctx.log, error: ctx.error, }) }
Function Environment Variables
Both functions that you just created require access to the environment variables below. You can set
them globally in your Appwrite project settings or scope them to each function. If you opted
for the scoped approach ensure the values match across functions. Also, ensure the config
does not change over time if you run the runMigrationSequence at least once. The code is not
adapted for configuration changes. While they are possible, we do not recommend doing them, unless
you have a good reason and planned a transition. This includes updating environment variables,
build paths, function names, or repository changes. Mistakes can leave your application in
inconsistent states.
# Required
APPWRITE_API_KEY=<your-appwrite-api-key>
# Required
APPWRITE_ENDPOINT=<your-appwrite-endpoint>
# Required (when running locally instead of a serverless function environment)
APPWRITE_FUNCTION_PROJECT_ID=<your-appwrite-project-id>
# Required
MIGRATIONS_DATABASE_ID=<database-id>
# Defaults to 'appwritedatabasemigrationtool'
MIGRATIONS_COLLECTION_ID=<collection-id>
# Defaults to 'Migrations'
MIGRATIONS_COLLECTION_NAME=<collection-name>
# Defaults to './migrations'
MIGRATIONS_HOME=<relative-path-to-folder-where-runner-finds-your-migrations>Finalize ADMT Setup
- Execute
Create Migrations Collectiononce and only once per environment/project.- We do prevent duplicate creations. 😇
- Check that the
Migrationscollection was created with at least the following attributes:$id: Stringapplied: Booleanname: Stringtimestamp: Integer
Create Your First Migration
Use our codegen tool to create a new Migration JavaScript file. We give you type annotations through JSDocs (works just like TypeScript) without needing you to do transpilation steps.
- The codegen tool is
NodeandBuncompatible. - Your description will be converted to
PascalCase.- Whitespaces are not allowed.
# E.g.: npx admt new-migration --outpath ./functions/database/migrations --descriptor CreateProductsCollection npx admt new-migration --outpath <relative-path> --descriptor <migration-summary>
- The codegen tool is
Use the
databaseServiceparameter ofupanddownto write your migration.- The parameter is an instance of
node-appwriteDatabases class.
- The parameter is an instance of
Once you are done, deploy push your changes through the environment pipelines.
- E.g.: Push to
stagingexecute theRun Migration Sequencefunction on Appwrite UI. Verify all is good. Finally push toproductionand run the sequence there.
- E.g.: Push to
Usage, Rules, Recommendations and, FAQ
Rules
- Migrations must complete within Appwrite Cloud defined timeout for the function (default is 15s).
- Longer migrations should be run from local maching, by exporting variables in your
.env.localfor example.
- Longer migrations should be run from local maching, by exporting variables in your
- Never change the file name of a migration file.
- Never change the class name of a migration class.
- Always use codegen tools to create new migration files or other supported operations.
Recommendations
- Avoid changing the contents of a migration that you have pushed to a production-like environment.
- Unless you can confidently revert the database state (e.g.: staging) without affecting end-users.
- Provide a meaningful descriptions for your migration using the
--descriptorflag.- Keep them short, like
git commitmessages.
- Keep them short, like
- Follow the expand-and-contract pattern.
- Read here.
- Follow the single-responsibility principle.
- We do not have direct access to Appwrite's MariaDB instances, thus no real transaction mechanisms are used.
- It's better to do incremental migrations then one migration that might leave your app in an inconsistent state. Plan your migrations!
- Avoid abstractions in your migration files. They are subject to change in the future. If you use them, ensure that whatever happens, the output of the migration up sequence is always the same. A change of output in a migration M may cause a migration M + x, x > 0, to no longer work as intended.
- Test your migration locally and your staging environment before releasing to production!
- Mistakes happen. If you pushed to production and applying 'down' is not possible, we recommend
creating a new migration file to patch the issue.
- If that is not possible, restore your database to a previous point-in-time (backup).
FAQ
How do I bundle my migrations in the final function bundle?
How you bundle your migrations depends on your overall language choices and how you choose to set up your Appwrite Function source repository structure. My personal setup using a Bun based functions has the following Function Configurations. Please note that appwrite does not allow you to do
newlinewith continuation markers\like I did in the example below example (for readability purposes). It expects the entire command to be written in one line.Entrypoint: dist/database/migrations-create-collection.js Build Settings: \ bun install --production \ && bun build ./functions/database/migrations-create-collection.ts --outdir ./dist/database \ && cp -r ./functions/database/migrations ./dist/database/migrations