Sqitch

sqitch-configuration

Name

sqitch-configuration - Hierarchical engine and target configuration

Description

The specification of database targets is core to Sqitch database change management. A target consists of a database connection URI, a plan file, change script directories, a registry schema or database name, and the path to a database engine command-line client. Sqitch determines the values for these attributes via a hierarchical evaluation of the runtime configuration, examining and selecting from these values:

  1. Command-line options
  2. Target-specific configuration
  3. Engine-specific configuration
  4. Core configuration
  5. A reasonable default

This document explains how this evaluation works, and how to use the init, config, engine, and target commands to configure these values for various deployment scenarios.

Project Initialization

Typically, the first thing you do with Sqitch is use the init command to start a new project. Now, the most important thing Sqitch needs to know is what database engine you’ll be managing, so it’s best to use --engine to configure the engine right up front to start off on the right foot. Here, we start a project called “widgets” to manage PostgreSQL databases:

> sqitch init widgets --engine pg
Created sqitch.conf
Created sqitch.plan
Created deploy/
Created revert/
Created verify/

This creates a very simple configuration file with most of the settings commented out, like so:

> cat sqitch.conf
[core]
  engine = pg
  # plan_file = sqitch.plan
  # top_dir = .
# [engine "pg"]
  # target = db:pg:
  # registry = sqitch
  # client = psql

The [core] section contains default configurations, the most important of which is the default engine, pg. Of course, it’s the only engine this project supports, and the values of the other configuration variables are reasonable for a single-engine project. If your Sqitch project never needs to manage more than one database engine, this might be all you need: the current directory is the top directory of the project, and it’s here you’ll find the plan file as well as the deploy, revert, and verify script directories. Once you start using the add command to add changes, and the deploy command to deploy changes to a database, these variables will be used extensively.

The [engine "pg"] section houses the variables specific to the engine. The target defines the default database URI for connecting to a PostgreSQL database. As you can see there isn’t much here, but if you were to distribute this project, it’s likely that your users would specify a target URI when deploying to their own databases. The registry determines where Sqitch will store its own metadata when managing a database; generally the default, “sqitch”, is fine.

More interesting, perhaps, is the client setting, which defaults to the appropriate engine-specific client name appropriate for your OS. In this example, sqitch will assume it can find psql in your path.

Global Configuration

But sometimes that’s not the case. Let’s say that the psql client on your system is not in the path, but instead in /usr/local/pgsql/bin/psql. You could set its location right here in the project configuration file, but that won’t do if you end up distributing the project to other users who might have their client somewhere else. For that use case, the default path-specific value is probably best.

A better idea is to tell Sqitch where to find psql for all of your projects. Use the config command’s --user option to set that configuration for yourself:

> sqitch config --user engine.pg.client /usr/local/pgsql/bin/psql

This won’t change the project configuration file at all, but add the value to ~/.sqitch/sqitch.conf, which is your personal cross-project Sqitch configuration. In other words, it sets the PostgreSQL client for all Sqitch projects you manage on this host. In fact, it can be a good idea to configure clients not in the path first thing whenever you start working on a new host:

> sqitch config --user user.name 'Marge N. O’Vera'
> sqitch config --user user.email 'marge@example.com'
> sqitch config --user engine.pg.client /usr/local/pgsql/bin/psql
> sqitch config --user engine.mysql.client /usr/local/mysql/bin/mysql
> sqitch config --user engine.sqlite.client /sbin/sqlite3

If you’d like to make the configuration global to all accounts on your host, use the --system option, instead:

> sudo sqitch config --system engine.pg.client /usr/local/pgsql/bin/psql
> sudo sqitch config --system engine.mysql.client /usr/local/mysql/bin/mysql
> sudo sqitch config --system engine.sqlite.client /sbin/sqlite3

That will put the values into the global Sqitch configuration file, which is in `sqitch --etc-path`/sqitch.conf.

Engine Configuration

So you’ve got the widgets project well developed, and now you’ve been asked to port it to SQLite. Fundamentally, that means porting all of your deploy, revert, and verify scripts. The simplest way to organize files for this configuration is with top-level directories for each engine. First, let’s move the existing PostgreSQL stuff to a subdirectory.

> mkdir pg
> mv deploy revert verify sqitch.plan pg
> ls  pg
deploy/ revert/ sqitch.plan verify/

Now we need to tell Sqitch where things are. To create an engine-specific configuration, use the engine command’s add action:

sqitch engine add pg --top-dir pg

The add action adds the pg engine to the configuration, setting the top directory to our newly-created pg directory. The configuration looks like this (with comments removed for clarity):

[core]
  engine = pg
[engine "pg"]
  target = db:pg:
  top_dir = pg

Curious about all the other settings for the engine? Let sqitch engine show show you:

> sqitch engine show pg
* pg
    Target:        db:pg:
    Registry:      sqitch
    Client:        psql
    Top Directory: pg
    Plan File:     pg/sqitch.plan
    Extension:     sql
    Script Directories:
      Deploy:      pg/deploy
      Revert:      pg/revert
      Verify:      pg/verify
    Reworked Script Directories:
      Reworked:    pg
      Deploy:      pg/deploy
      Revert:      pg/revert
      Verify:      pg/verify
    No Variables

The show action nicely presents the result of the fully-evaluated configuration, even though only the top directory and client have been set. Nice, right?

Now, to add the SQLite support. There are two basic ways to go about it. We’ll start with the more obvious one.

Separate Plans

The first approach is to create an entirely independent SQLite project with its own plan and scripts. This is almost like starting from scratch: just create a new directory and add the Sqitch engine using it for its top directory: add initialize it as a new Sqitch project:

> sqitch engine add sqlite --top-dir sqlite
Created sqlite/
Created sqlite/sqitch.plan
Created sqlite/deploy/
Created sqlite/revert/
Created sqlite/verify/

Note the creation of a new sqlite/sqitch.conf file. It will have copied the project name and URI from the existing plan file. The SQLite configuration is now added to the configuration file:

> sqitch engine show sqlite
* sqlite
    Target:        db:sqlite:
    Registry:      sqitch
    Client:        sqlite3
    Top Directory: sqlite
    Plan File:     sqlite/sqitch.plan
    Extension:     sql
    Script Directories:
      Deploy:      sqlite/deploy
      Revert:      sqlite/revert
      Verify:      sqlite/verify
    Reworked Script Directories:
      Reworked:    sqlite
      Deploy:      sqlite/deploy
      Revert:      sqlite/revert
      Verify:      sqlite/verify
    No Variables

Good, everything’s in the right place. Start adding changes to the SQLite plan by passing the engine name to the add command:

> sqitch add users sqlite -m 'Creates users table.'
Created sqlite/deploy/users.sql
Created sqlite/revert/users.sql
Created sqlite/verify/users.sql
Added "users" to sqlite/sqitch.plan

Pass pg when adding PostgreSQL changes, or omit it, in which case Sqitch will fall back on the default engine, defined by the core.engine variable set when we created the PostgreSQL project. Want to add a change with the same name to both engines? Simply pass them both, or use the --all option:

> sqitch add users --all -m 'Creates users table.'
Created pg/deploy/users.sql
Created pg/revert/users.sql
Created pg/test/users.sql
Created pg/verify/users.sql
Added "users" to pg/sqitch.plan
Created sqlite/deploy/users.sql
Created sqlite/revert/users.sql
Created sqlite/test/users.sql
Created sqlite/verify/users.sql
Added "users" to sqlite/sqitch.plan

Shared Plan

The other approach is to have both the PostgreSQL and the SQLite projects share the same plan. In that case, we should move the plan file out of the PostgreSQL directory:

> mv pg/sqitch.plan .
> sqitch engine alter pg --plan-file sqitch.plan
> sqitch engine show pg
* pg
    Target:        db:pg:
    Registry:      sqitch
    Client:        psql
    Top Directory: pg
    Plan File:     sqitch.plan
    Extension:     sql
    Script Directories:
      Deploy:      pg/deploy
      Revert:      pg/revert
      Verify:      pg/verify
    Reworked Script Directories:
      Reworked:    pg
      Deploy:      pg/deploy
      Revert:      pg/revert
      Verify:      pg/verify
    No Variables

Good, it’s now using ./sqitch.plan. Now let’s start the SQLite project. Since we’re going to use the same plan, we’ll need to port all the scripts from PostgreSQL. Let’s just copy them, and then configure the SQLite engine to use the shared plan file:

> cp -rf pg sqlite
> sqitch engine add sqlite --plan-file sqitch.plan --top-dir sqlite
> sqitch engine show sqlite
* sqlite
    Target:           db:sqlite:
    Registry:         sqitch
    Client:           sqlite3
    Top Directory:    sqlite
    Plan File:        sqitch.plan
    Extension:        sql
    Script Directories:
      Deploy:      sqlite/deploy
      Revert:      sqlite/revert
      Verify:      sqlite/verify
    Reworked Script Directories:
      Reworked:    sqlite
      Deploy:      sqlite/deploy
      Revert:      sqlite/revert
      Verify:      sqlite/verify
    No Variables

Looks good! Now port all the scripts in the sqlite directory from PostgreSQL to SQLite and you’re ready to go.

Later, when you want to add a new change to both projects, just pass the --all option to the add command:

> sqitch add users --all -n 'Creates users table.'
Created pg/deploy/users.sql
Created pg/revert/users.sql
Created pg/verify/users.sql
Created sqlite/deploy/users.sql
Created sqlite/revert/users.sql
Created sqlite/verify/users.sql
Added "users" to sqitch.plan

This option also works for the tag, rework, and bundle commands. If you know you always want to act on all plans, set the all configuration variable for each command:

sqitch config --bool add.all 1
sqitch config --bool tag.all 1
sqitch config --bool rework.all 1
sqitch config --bool bundle.all 1

Database Interactions

With either of these two approaches, you can manage database interactions by passing an engine name or a database URI to the database commands. For example, to deploy to a PostgreSQL database to the default PostgreSQL database:

sqitch deploy pg

You usually won’t want to use the default database in production, though. Here’s how to deploy to a PostgreSQL database named “widgets” on host db.example.com:

sqitch deploy db:pg://db.example.com/widgets

Sqitch is smart enough to pick out the proper engine from the URI. If you pass a db:pg: URI, rest assured that Sqitch won’t try to deploy the SQLite changes. Use a db:sqlite: URI to interact with an SQLite database:

sqitch log db:sqlite:/var/db/widgets.db

The commands that take engine and target URI arguments include:

Target Configuration

Great, now we can easily manage changes for multiple database engines. But what about multiple databases for the same engine? For example, you might want to deploy your database to two hosts in a primary/standby configuration. To make things as simple as possible for your IT organization, set up named targets for those servers:

> sqitch target add prod-primary db:pg://sqitch@db1.example.com/widgets
> sqitch target add prod-standby db:pg://sqitch@db2.example.com/widgets

Targets inherit configuration from engines, based on the engine specified in the URI. Thus the configuration all comes together:

> sqitch target show prod-primary prod-standby
* prod-primary
    URI:           db:pg://sqitch@db1.example.com/widgets
    Registry:      sqitch
    Client:        psql
    Top Directory: pg
    Plan File:     sqitch.plan
    Extension:     sql
    Script Directories:
      Deploy:      pg/deploy
      Revert:      pg/revert
      Verify:      pg/verify
    Reworked Script Directories:
      Reworked:    pg
      Deploy:      pg/deploy
      Revert:      pg/revert
      Verify:      pg/verify
    No Variables
* prod-standby
    URI:           db:pg://sqitch@db2.example.com/widgets
    Registry:      sqitch
    Client:        psql
    Top Directory: pg
    Plan File:     sqitch.plan
    Extension:     sql
    Script Directories:
      Deploy:      pg/deploy
      Revert:      pg/revert
      Verify:      pg/verify
    Reworked Script Directories:
      Reworked:    pg
      Deploy:      pg/deploy
      Revert:      pg/revert
      Verify:      pg/verify
    No Variables

Note the use of the shared plan and the pg directory for scripts. We can add a target for our SQLite database, too. Maybe it’s used for development?

> sqitch target add dev-sqlite db:sqlite:/var/db/widgets_dev.db
> sqitch target show dev-sqlite
* dev-sqlite
    URI:           db:sqlite:/var/db/widgets_dev.db
    Registry:      sqitch
    Client:        sqlite3
    Top Directory: sqlite
    Plan File:     sqitch.plan
    Extension:     sql
    Script Directories:
      Deploy:      sqlite/deploy
      Revert:      sqlite/revert
      Verify:      sqlite/verify
    Reworked Script Directories:
      Reworked:    sqlite
      Deploy:      sqlite/deploy
      Revert:      sqlite/revert
      Verify:      sqlite/verify
    No Variables

Now deploying any of these databases is as simple as specifying the target name when executing the deploy command (assuming the sqitch user is configured to authenticate to PostgreSQL without prompting for a password):

> sqitch deploy prod-primary
> sqitch deploy prod-standby

Want them all? Just query the targets and pass each in turn:

for target in `sqitch target | grep prod-`; do
    sqitch deploy $target
done

The commands that accept a target name are identical to those that take an engine name or target URI, as described in “Database Interactions”.

Different Target, Different Plan

What about a project that manages different – but related – schemas on the same engine? For example, say you have two plans for PostgreSQL, one for a canonical data store, and one for a read-only copy that will have a subset of data replicated to it. Maybe your billing database just needs an up-to-date copy of the customers and users tables.

Targets can help us here, too. Just create the new plan file. It might use some of the same change scripts as the canonical plan, or its own scripts, or some of each. Just be sure all of its scripts are in the same top directory. Then add targets for the specific servers and plans:

> sqitch target add prod-primary db:pg://db1.example.com/widgets
> sqitch target add prod-billing db:pg://cpa.example.com/billing --plan-file target.plan
> sqitch target show prod-billing
* prod-billing
    URI:           db:pg://cpa.example.com/billing
    Registry:      sqitch
    Client:        psql
    Top Directory: pg
    Plan File:     target.plan
    Extension:     sql
    Script Directories:
      Deploy:      pg/deploy
      Revert:      pg/revert
      Verify:      pg/verify
    Reworked Script Directories:
      Reworked:    pg
      Deploy:      pg/deploy
      Revert:      pg/revert
      Verify:      pg/verify
    No Variables

Now, any management of the prod-billing target will use the target.plan plan file. Want to add changes to that plan? specify the plan file. Here’s an example that re-uses the existing change scripts:

> sqitch add users target.plan -n 'Creates users table.'
Skipped pg/deploy/users.sql: already exists
Skipped pg/revert/users.sql: already exists
Skipped pg/test/users.sql: already exists
Skipped pg/verify/users.sql: already exists
Added "users" to target.plan

Overworked

Say you’ve been working on your project for some time, and now you have a slew of changes you’ve reworked. (You really only do that with procedures and views, right? Because it’s silly to use for ALTER statements; just add new changes in those cases.) As a result, your deploy, revert, and verify directories are full of files representing older versions of the changes, all containing the @ symbol, and they’re starting to get in the way (in general you’ll never modify them). Here’s an example adapted from a real project:

> find pg -name '*@*'
pg/deploy/extensions@v2.9.0.sql
pg/deploy/jobs/func_enabler@v2.6.1.sql
pg/deploy/stem/func_check_all_widgets@v2.11.0.sql
pg/deploy/stem/func_check_all_widgets@v2.12.2.sql
pg/deploy/stem/func_check_all_widgets@v2.12.3.sql
pg/deploy/crank/func_update_jobs@v2.12.0.sql
pg/deploy/crank/func_update_jobs@v2.8.0.sql
pg/deploy/utility/func_get_sleepercell@v2.9.0.sql
pg/deploy/utility/func_update_connection@v2.10.0.sql
pg/deploy/utility/func_update_connection@v2.10.1.sql
pg/deploy/utility/func_update_connection@v2.11.0.sql
pg/revert/extensions@v2.9.0.sql
pg/revert/jobs/func_enabler@v2.6.1.sql
pg/revert/stem/func_check_all_widgets@v2.11.0.sql
pg/revert/stem/func_check_all_widgets@v2.12.2.sql
pg/revert/stem/func_check_all_widgets@v2.12.3.sql
pg/revert/crank/func_update_jobs@v2.12.0.sql
pg/revert/crank/func_update_jobs@v2.8.0.sql
pg/revert/utility/func_get_sleepercell@v2.9.0.sql
pg/revert/utility/func_update_connection@v2.10.0.sql
pg/revert/utility/func_update_connection@v2.10.1.sql
pg/revert/utility/func_update_connection@v2.11.0.sql
pg/verify/extensions@v2.9.0.sql
pg/verify/jobs/func_enabler@v2.6.1.sql
pg/verify/stem/func_check_all_widgets@v2.11.0.sql
pg/verify/stem/func_check_all_widgets@v2.12.2.sql
pg/verify/stem/func_check_all_widgets@v2.12.3.sql
pg/verify/crank/func_update_jobs@v2.12.0.sql
pg/verify/crank/func_update_jobs@v2.8.0.sql
pg/verify/utility/func_get_sleepercell@v2.9.0.sql
pg/verify/utility/func_update_connection@v2.10.0.sql
pg/verify/utility/func_update_connection@v2.10.1.sql
pg/verify/utility/func_update_connection@v2.11.0.sql

Ugh. Wouldn’t it be nice to move them out of the way? Of course it would! So let’s do that. We want all of the PostgreSQL engine’s reworked scripts all to go into to a new directory named “reworked”, so tell Sqitch where to find them:

> sqitch engine alter pg --dir reworked=pg/reworked
Created pg/reworked/deploy/
Created pg/reworked/revert/
Created pg/reworked/verify/

Great, it created the new directories. Note that if you wanted the directories to have different names or locations, you can use the reworked_deploy, reworked_revert, and reworked_verify options.

Now all we have to do is move the files:

cd pg
for file in `find . -name '*@*'`
do
    mkdir -p reworked/`dirname $file`
    mv $file reworked/`dirname $file`
done
cd ..

Now all the reworked deploy files are in pg/reworked/deploy, the reworked revert files are in pg/reworked/revert, and the reworked verify files are in pg/reworked/verify. And you’re good to go! From here on in Sqitch always knows to find the reworked scripts when doing a deploy, revert, or bundle. And meanwhile, they’re tucked out of the way, less likely to break your brain or your IDE.

Other Options

You can see by the output of the init, engine, and target commands that there are quite a few other properties that can be set on a per-engine or per-target database. To determine the value of each, Sqitch looks at a combination of command-line options and configuration variables. Here’s a complete list, including specification of their values and how to set them.

See Also

Sqitch

Part of the sqitch suite.