Backup & Restore Metadata

Introduction

The goal of OpenMetadata is to enable company-wide collaboration around metadata. The more we use it, the more value this brings to the table, which means that keeping the metadata safe can become a critical activity for our Disaster Recovery practices.

While there are cloud services that feature automatic snapshots and replication, the metadata CLI now allows all users to perform backups regardless of the underlying infrastructure.

Requirements

The backup CLI needs to be used with openmetadata-ingestion version 0.13.1 or higher.

However, we recommend to make sure that you are always in the latest openmetadata-ingestion version to have all the improvements shipped in the CLI.

Installation

The CLI comes bundled in the base openmetadata-ingestion Python package. You can install it with:

One of the backup features is to upload the generated backup to cloud storage (currently supporting S3 and Azure Blob). To use this, you can instead install the package with the backup plugin:

Requirements & Considerations

This is a custom utility. As almost all tables contain GENERATED columns, directly using mysqldump is not an option out of the box, as it would require some further cleaning steps to get the data right.

Instead, we have created a utility that will just dump the necessary data.

The requirement for running the process is that the target database should have the Flyway migrations executed.

The backup utility will provide an SQL file which will do two things:

1. TRUNCATE the OpenMetadata tables
2. INSERT the data that has been saved

You can then run the script's statements to restore the data.

Backup CLI

After the installation, we can take a look at the different options to run the CLI:

If using MySQL:

If using Postgres, add the -s parameter specifying the schema:

Database Connection

There is a set of four required parameters, the minimum required for us to access the database service and run the backup: host, user, password and database to point to. Note that the user should have at least read access to the database. By default, we'll try to connect through the port 3306, but this can be overridden with the --port option.

Output

The CLI will create a dump file that looks like openmetadata_YYYYmmddHHMM_backup.sql. This will help us identify the date each backup was generated. We can also specify an output path, which we'll create if it does not exist, via --output.

Uploading to S3

To run this, make sure to have AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY as environment variables with permissions to the bucket that you'd like to point to. Afterwards, we can just use --upload <endpoint> <bucket> <key> to have the CLI upload the file. In this case, you'll get both the local dump file and the one in the cloud.

Uploading to Azure Blob

To run this, make sure to have Azure CLI configured with permissions to the Blob that you'd like to point to. Afterwards, we can just use --upload <account_url> <container> <folder> to have the CLI upload the file. In this case, you'll get both the local dump file and the one in the cloud.

Connection Options and Arguments

You can pass any required connection options or arguments to the MySQL connection via -o <opt1>, -o <opt2> [...] or -a <arg1>, -a <arg2> [...].

Trying it out

We can do a test locally preparing some containers:

1. sh docker/run_local_docker.sh to start the docker compose service.
2. docker run -p 9000:9000 -p 9001:9001 minio/minio server /data --console-address ":9001" to start minio, an object storage S3 compatible.
3. Connect to http://localhost:9001 URL to reach the minio console and create a bucket called my-bucket
4. Finally, we just need to prepare the environment variables as:

An example of S3 CLI call will look as:

And we'll get the following output:

If we now head to the minio console and check the my-backup bucket, we'll see our SQL dump in there.

An example of Azure Blob CLI call will look as:

And we'll get the following output:

Restore Metadata

Make sure that when restoring metadata, your OpenMetadata server is NOT RUNNING. Otherwise, there could be clashes on cached values and specific elements that the server creates at start-time and that will be present on the restore SQL file as well.

Introduction

SQL file which is generated using Backup metadata CLI can restore using Restore metadata CLI.

Requirements

The restore CLI needs to be used with openmetadata-ingestion version 0.12.1 or higher.

Restore CLI

After the installation, we can take a look at the different options to run the CLI:

Output

The CLI will give messages like this Backup restored from openmetadata_202209301715_backup.sql when backup restored completed.

Trying it out

An example CLI call will look as:

And we'll get the following output: