Home - News ( United States | United Kingdom | Italy | Germany ) - Football scores

Showing content from https://cloud.google.com/spanner/docs/migrating-postgres-spanner below:

Migrate from PostgreSQL to Spanner (GoogleSQL dialect)

This page provides guidance on migrating an open source PostgreSQL database to Spanner.

Migration involves the following tasks:

• Mapping a PostgreSQL schema to a Spanner schema.
• Creating a Spanner instance, database, and schema.
• Refactoring the application to work with your Spanner database.
• Migrating your data.
• Verifying the new system and moving it to production status.

This page also provides some example schemas using tables from the MusicBrainz PostgreSQL database.

Your first step in moving a database from PostgreSQL to Spanner is to determine what schema changes you must make. Use pg_dump to create Data Definition Language (DDL) statements that define the objects in your PostgreSQL database, and then modify the statements as described in the following sections. After you update the DDL statements, use them to create your database in a Spanner instance.

The following table describes how PostgreSQL data types map to Spanner data types. Update the data types in your DDL statements from PostgreSQL data types to Spanner data types.

The Spanner NUMERIC data type supports up to 38 digits of precision and 9 decimal digits of scale.

If you require greater precision, see Storing arbitrary precision numeric data for alternative mechanisms.

For tables in your Spanner database that you frequently append to, avoid using primary keys that monotonically increase or decrease, as this approach causes hotspots during writes. Instead, modify the DDL CREATE TABLE statements so that they use supported primary key strategies. If you are using a PostgreSQL feature such as a UUID data type or function, SERIAL data types, IDENTITY column, or sequence, you can use the auto-generated key migration strategies that we recommend.

Note that after you designate your primary key, you can't add or remove a primary key column, or change a primary key value later without deleting and recreating the table. For more information on how to designate your primary key, see Schema and data model - primary keys.

During migration, you might need to keep some existing monotonically increasing integer keys. If you need to keep these kinds of keys on a frequently updated table with a lot of operations on these keys, you can avoid creating hotspots by prefixing the existing key with a pseudo-random number. This technique causes Spanner to redistribute the rows. See What DBAs need to know about Spanner, part 1: Keys and indexes for more information on using this approach.

Learn about foreign keys support in Spanner.

PostgreSQL b-tree indexes are similar to secondary indexes in Spanner. In a Spanner database you use secondary indexes to index commonly searched columns for better performance, and to replace any UNIQUE constraints specified in your tables. For example, if your PostgreSQL DDL has this statement:

   CREATE TABLE customer (
      id CHAR (5) PRIMARY KEY,
      first_name VARCHAR (50),
      last_name VARCHAR (50),
      email VARCHAR (50) UNIQUE
     );

You would use this statement in your Spanner DDL:

   CREATE TABLE customer (
      id STRING(5),
      first_name STRING(50),
      last_name STRING(50),
      email STRING(50)
      ) PRIMARY KEY (id);

    CREATE UNIQUE INDEX customer_emails ON customer(email);

You can find the indexes for any of your PostgreSQL tables by running the \di meta-command in psql.

After you determine the indexes that you need, add CREATE INDEX statements to create them. Follow the guidance at Creating indexes.

Spanner implements indexes as tables, so indexing monotonically increasing columns (like those containing TIMESTAMP data) can cause a hotspot. See What DBAs need to know about Spanner, part 1: Keys and indexes for more information on methods to avoid hotspots.

Learn about CHECK constraint support in Spanner.

You must create the functionality of the following objects in your application logic:

• Views
• Triggers
• Stored procedures
• User-defined functions (UDFs)
• Columns that use serial data types as sequence generators

Keep the following tips in mind when migrating this functionality into application logic:

• You must migrate any SQL statements that you use from the PostgreSQL SQL dialect to the GoogleSQL dialect.
• If you use cursors, you can rework the query to use offsets and limits.

After you update your DDL statements to conform to Spanner schema requirements, use it to create your database in Spanner.

1. Create a Spanner instance. Follow the guidance in Instances to determine the correct regional configuration and compute capacity to support your performance goals.

2. Create the database by using either the Google Cloud console or the gcloud command-line tool:

1. Go to the instances page
2. Click on the name of the instance that you want to create the example database in to open the Instance details page.
3. Click Create Database.
4. Type a name for the database and click Continue.
5. In the Define your database schema section, toggle the Edit as text control.
6. Copy and paste your DDL statements into the DDL statements field.
7. Click Create.

1. Install the gcloud CLI.
2. Use the gcloud spanner databases create command to create the database:

   gcloud spanner databases create DATABASE_NAME --instance=INSTANCE_NAME
   --ddl='DDL1' --ddl='DDL2'

• DATABASE_NAME is the name of your database.
• INSTANCE_NAME is the Spanner instance that you created.
• DDLn are your modified DDL statements.

After you create the database, follow the instructions in Apply IAM roles to create user accounts and grant permissions to the Spanner instance and database.

In addition to the code needed to replace the preceding database objects, you must add application logic to handle the following functionality:

• Hashing primary keys for writes, for tables that have high write rates to sequential keys.
• Validating data, not already covered by CHECK constraints.
• Referential integrity checks not already covered by foreign keys, table interleaving or application logic, including functionality handled by triggers in the PostgreSQL schema.

We recommend using the following process when refactoring:

1. Find all of your application code that accesses the database, and refactor it into a single module or library. That way, you know exactly what code accesses to the database, and therefore exactly what code needs to be modified.
2. Write code that performs reads and writes on the Spanner instance, providing parallel functionality to the original code that reads and writes to PostgreSQL. During writes, update the entire row, not just the columns that have been changed, to ensure that the data in Spanner is identical to that in PostgreSQL.
3. Write code that replaces the functionality of the database objects and functions that aren't available in Spanner.

After you create your Spanner database and refactor your application code, you can migrate your data to Spanner.

1. Use the PostgreSQL COPY command to dump data to .csv files.

2. Upload the .csv files to Cloud Storage.

   1. Create a Cloud Storage bucket.
   2. In the Cloud Storage console, click on the bucket name to open the bucket browser.
   3. Click Upload Files.
   4. Navigate to the directory containing the .csv files and select them.
   5. Click Open.

3. Create an application to import data into Spanner. This application could use Dataflow or it could use the client libraries directly. Make sure to follow the guidance in Bulk data loading best practices to get the best performance.

Test all application functions against the Spanner instance to verify that they work as expected. Run production-level workloads to ensure the performance meets your needs. Update the compute capacity as needed to meet your performance goals.

After you complete the initial application testing, turn up the new system using one of the following processes. Offline migration is the simplest way to migrate. However, this approach makes your application unavailable for a period of time, and it provides no rollback path if you find data issues later on. To perform an offline migration:

1. Delete all the data in the Spanner database.
2. Shut down the application that targets the PostgreSQL database.
3. Export all data from the PostgreSQL database and import it into the Spanner database as described in Migration overview.

4. Start up the application that targets the Spanner database.

Live migration is possible and requires extensive changes to your application to support the migration.

These examples show the CREATE TABLE statements for several tables in the MusicBrainz PostgreSQL database schema. Each example includes both the PostgreSQL schema and the Spanner schema.

CREATE TABLE artist_credit (
 hashed_id STRING(4),
 id INT64,
 name STRING(MAX) NOT NULL,
 artist_count INT64 NOT NULL,
 ref_count INT64,
 created TIMESTAMP OPTIONS (
    allow_commit_timestamp = true
 ),
) PRIMARY KEY(hashed_id, id);

CREATE TABLE artist_credit (
 id SERIAL,
 name VARCHAR NOT NULL,
 artist_count SMALLINT NOT NULL,
 ref_count INTEGER DEFAULT 0,
 created TIMESTAMP WITH TIME ZONE DEFAULT NOW()
);

CREATE TABLE recording (
  hashed_id STRING(36),
  id INT64,
  gid STRING(36) NOT NULL,
  name STRING(MAX) NOT NULL,
  artist_credit_hid STRING(36) NOT NULL,
  artist_credit_id INT64 NOT NULL,
  length INT64,
  comment STRING(255) NOT NULL,
  edits_pending INT64 NOT NULL,
  last_updated TIMESTAMP OPTIONS (
     allow_commit_timestamp = true
  ),
  video BOOL NOT NULL,
) PRIMARY KEY(hashed_id, id);

CREATE TABLE recording (
  id SERIAL,
  gid UUID NOT NULL,
  name VARCHAR NOT NULL,
  artist_credit INTEGER NOT NULL, -- references artist_credit.id
  length INTEGER CHECK (length IS NULL OR length > 0),
  comment VARCHAR(255) NOT NULL DEFAULT '',
  edits_pending INTEGER NOT NULL DEFAULT 0 CHECK (edits_pending >= 0),
  last_updated TIMESTAMP WITH TIME ZONE DEFAULT NOW(),
  video BOOLEAN NOT NULL DEFAULT FALSE
);

CREATE TABLE recording_alias (
  hashed_id STRING(36)  NOT NULL,
  id INT64  NOT NULL,
  alias_id INT64,
  name STRING(MAX)  NOT NULL,
  locale STRING(MAX),
  edits_pending INT64  NOT NULL,
  last_updated TIMESTAMP NOT NULL OPTIONS (
     allow_commit_timestamp = true
  ),
  type INT64,
  sort_name STRING(MAX)  NOT NULL,
  begin_date_year INT64,
  begin_date_month INT64,
  begin_date_day INT64,
  end_date_year INT64,
  end_date_month INT64,
  end_date_day INT64,
  primary_for_locale BOOL NOT NULL,
  ended BOOL NOT NULL,
) PRIMARY KEY(hashed_id, id, alias_id),
 INTERLEAVE IN PARENT recording ON DELETE NO ACTION;

CREATE TABLE recording_alias (
  id SERIAL, --PK
  recording INTEGER NOT NULL, -- references recording.id
  name VARCHAR NOT NULL,
  locale TEXT,
  edits_pending INTEGER NOT NULL DEFAULT 0 CHECK (edits_pending >=0),
  last_updated TIMESTAMP WITH TIME ZONE DEFAULT NOW(),
  type INTEGER, -- references recording_alias_type.id
  sort_name VARCHAR NOT NULL,
  begin_date_year SMALLINT,
  begin_date_month SMALLINT,
  begin_date_day SMALLINT,
  end_date_year SMALLINT,
  end_date_month SMALLINT,
  end_date_day SMALLINT,
  primary_for_locale BOOLEAN NOT NULL DEFAULT false,
  ended BOOLEAN NOT NULL DEFAULT FALSE
  -- CHECK constraint skipped for brevity
);

RetroSearch is an open source project built by @garambo | Open a GitHub Issue

Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo

HTML: 3.2 | Encoding: UTF-8 | Version: 0.7.4