gem-hunter

Singer tap for Postgres.

Built with the Meltano Singer SDK.

• catalog
• state
• discover
• about
• stream-maps
• schema-flattening

This tap is tested with all actively supported Python and PostgreSQL versions. At the time of writing, this includes Python 3.10 through 3.13 and PostgreSQL 14 through 18.

A full list of supported settings and capabilities is available by running: tap-postgres --about

This Singer tap will automatically import any environment variables within the working directory's .env if the --config=ENV is provided, such that config values will be considered if a matching environment variable is set either in the terminal context or in the .env file.

uv tool install meltanolabs-tap-postgres

You can easily run tap-postgres by itself or in a pipeline using Meltano.

tap-postgres --version
tap-postgres --help
tap-postgres --config CONFIG --discover > ./catalog.json

Prerequisites:

• uv

uv tool install --with tox-uv tox
uv tool install pre-commit
uv sync

We have set the provided keys in the .ssl directory to be valid for multiple centuries. However, we have also provided configuration instructions below to create all of the necessary files for testing SSL.

A list of each file and its purpose:

1. ca.crt: CA for client's certificate (stored on the server)
2. cert.crt: Client's certificate (stored on the client)
3. pkey.key: Client's private key (stored on the client)
4. public_pkey.key: Client's private key with incorrect file permissions (stored on the client)
5. root.crt: CA for server's certificate (stored on the client)
6. server.crt: Server's certificate (stored on the server)
7. server.key: Server's private key (stored on the server)

Run the following command to generate all relevant SSL files, with certificates valid for two centuries (73048 days).

chmod 0600 ssl/*.key
openssl req -new -x509 -days 73048 -nodes -out ssl/server.crt -keyout ssl/server.key -subj "/CN=localhost" &&
openssl req -new -x509 -days 73048 -nodes -out ssl/cert.crt -keyout ssl/pkey.key -subj "/CN=postgres" &&
cp ssl/server.crt ssl/root.crt &&
cp ssl/cert.crt ssl/ca.crt &&
cp ssl/pkey.key ssl/public_pkey.key &&
chown 999:999 ssl/server.key &&
chmod 600 ssl/server.key &&
chmod 600 ssl/pkey.key &&
chmod 644 ssl/public_pkey.key

Now that all of the SSL files have been set up, you're ready to set up tests with pytest.

Create tests within the tap_postgres/tests subfolder and then run:

uv run pytest

NOTE: Running the tests requires a locally running postgres. See tests/settings.py for the expected configuration.

You can also test the tap-postgres CLI interface directly using uv run:

uv run tap-postgres --help

Note: This tap will work in any Singer environment and does not require Meltano. Examples here are for convenience and to streamline end-to-end orchestration scenarios.

Your project comes with a custom meltano.yml project file already created.

Next, install Meltano (if you haven't already) and any needed plugins:

# Install meltano
uv tool install meltano
# Initialize meltano within this directory
cd tap-postgres
meltano install

Now you can test and orchestrate using Meltano:

# Test invocation:
meltano invoke tap-postgres --version
# OR run a test `elt` pipeline:
meltano elt tap-postgres target-jsonl

See the dev guide for more instructions on how to use the SDK to develop your own taps and targets.

This tap supports connecting to a Postgres database via an SSH tunnel (also known as a bastion server). This is useful if you need to connect to a database that is not publicly accessible. This is the same as using ssh -L, but this is done inside the tap itself.

An SSH tunnel is a method to securely forward network traffic. It uses the SSH protocol to encapsulate other protocols like HTTP, MySQL, Postgres, etc. This is particularly useful in scenarios where you need to access a service that is behind a firewall or in a network that you can't reach directly. In the context of this tap, you can use an SSH tunnel to access a Postgres database that's not accessible to the wider internet.

Here's a basic illustration of how an SSH tunnel works:

+-------------+             +-------------+             +-------------+
|    Local    | SSH tunnel  |   Bastion   |   Direct    |  Postgres   |
|   Machine   | <=========> |   Server    | <=========> |     DB      |
+-------------+ (encrypted) +-------------+ (unsecured) +-------------+

1. Local Machine: This is wherever this tap is running from, where you initiate the SSH tunnel. It's also referred to as the SSH client.
2. Bastion Server: This is a secure server that you have SSH access to, and that can connect to the remote server. All traffic in the SSH tunnel between your local machine and the bastion server is encrypted.
3. Remote Server: This is the server you want to connect to, in this case a PostgreSQL server. The connection between the bastion server and the remote server is a normal, potentially unsecured connection. However, because the bastion server is trusted, and because all traffic between your local machine and the bastion server is encrypted, you can safely transmit data to and from the remote server.

Setup

1. Ensure your bastion server is online.
2. Ensure you bastion server can access your Postgres database.
3. Have some method of accessing your bastion server. This could either be through password-based SSH authentication or through a hardwired connection to the server.
4. Install ssh-keygen on your client machine if it is not already installed.

Creating Keys

1. Run the command ssh-keygen.
   1. Enter the directory where you would like to save your key. If you're in doubt, the default directory is probably fine.
      • If you get a message similar to the one below asking if you wish to overwrite a previous key, enter n, then rerun ssh-keygen and manually specify the output_keyfile using the -f flag.

        /root/.ssh/id_rsa already exists.
        Overwrite (y/n)?
        

   2. If you wish, enter a passphrase to provide additional protection to your private key. SSH-based authentication is generally considered secure even without a passphrase, but a passphrase can provide an additional layer of security.
   3. You should now be presented with a message similar to the following, as well as a key fingerprint and ascii randomart image.

      Your identification has been saved in /root/.ssh/id_rsa
      Your public key has been saved in /root/.ssh/id_rsa.pub
      

2. Navigate to the indicated directory and find the two keys that were just generated. The file named id_rsa is your private key. Keep it safe. The file named id_rsa.pub is your public key, and needs to be transferred to your bastion server for your private key to be used.

Copying Keys

1. Now that you have a pair of keys, the public key needs to be transferred to your bastion server.
2. If you already have password-based SSH authentication configured, you can use the command ssh-copy-id [user]@[host] to copy your public key to the bastion server. Then you can move on to using your keys
3. If not, you'll need some other way to access your bastion server. Once you've accessed it, copy the id_rsa.pub file onto the bastion server in the ~/.ssh/authorized_keys file. You could do this using a tool such as rsync or with a cloud-based service.
   • Keep in mind: it's okay if your public key is exposed to the internet through a file-share or something similar. It is useless without your private key.

To connect through SSH, you will need to determine the following pieces of information. If you're missing something, go back to the section on Obtaining Keys to gather all the relevant information.

• The connection details for your Postgres database, the same as any other tap-postgres run. This includes host, port, username, password and database.
  • Alternatively, provide an sqlalchemy url. Keep in mind that many other configuration options are ignored when an sqlalchemy url is set, and ideally you should be able to accomplish everything through other configuration options. Consider making an issue in the tap-postrges repository if you find a reasonable use case that is unsupported by current configuration options.
  • Note that when your connection details are used, it will be from the perspective of the bastion server. This could change the meaning of local IP address or keywords such as "localhost".
• The hostname or ip address of the bastion server, provided in the ssh.host configuration option.
• The port for use with the bastion server, provided in the ssh.port configuration option.
• The username for authentication with the bastion server, provided in the ssh.username configuration option. This will require you to have setup an SSH login with the Bastion server.
• The private key you use for authentication with the bastion server, provided in the ssh.private_key configuration option. If your private key is protected by a password (alternatively called a "private key passphrase"), provide it in the ssh.private_key_password configuration option. If your private key doesn't have a password, you can safely leave this field blank.

After everything has been configured, be sure to indicate your use of an ssh tunnel to the tap by configuring the ssh.enable configuration option to be True. Then, you should be able to connect to your privately accessible Postgres database through the bastion server.

Log-based replication is an alternative to full-table and incremental syncs and syncs all changes to the database, including deletes. This feature is built based on postgres replication slots.

1. Managing replication slots - Log-based replication has to be set up and maintained on the database. This tap attempts to abstract away as much complexity as possible, but there's still potentially manual effort needed
2. Log Files - When a replication slot is setup the file that holds these logs will continue to grow until consumed, this can cause issues if the tap doesn't ingest these quickly enough due to outages, etc.

If and when someone finds more, please add them to this list!

Log-based replication will modify the schemas output by the tap. Specifically, all fields will be made nullable and non-required. The reason for this is that when the tap sends a message indicating that a record has been deleted, that message will leave all fields for that record (except primary keys) as null. The stream's schema must be capable of accommodating these messages, even if a source field in the database is not nullable. As a result, log-based schemas will have all fields nullable.

Note that changing what streams are selected after already beginning log-based replication can have unexpected consequences. To ensure consistent output, it is best to keep selected streams the same across invocations of the tap.

When a stream has no prior bookmark (i.e. it is new to log-based replication), the tap starts reading the replication slot from LSN 0 — the oldest WAL position the slot still retains. This means the tap will walk through all retained WAL for every registered table before reaching the current tip, which can be very slow if the slot has accumulated a large backlog.

Other CDC tools (e.g. Debezium, Airbyte) handle this by taking an initial snapshot of the new table at a known LSN and then starting WAL streaming from that point. This tap does not implement that pattern.

The recommended workaround is:

1. Sync the new table once using FULL_TABLE or INCREMENTAL replication to establish its baseline state.
2. Switch that stream's replication method to LOG_BASED for subsequent syncs.

This ensures the new stream has a valid starting LSN bookmark and does not force the other streams to re-read old WAL.

Note also that using log-based replication will cause the replication key for all streams to be set to "_sdc_lsn", which is the Postgres LSN for the record in question.

1. Ensure you are using PostgresSQL 9.4 or higher.

2. Need to access the master postgres instance

3. Install the wal2json plugin for your database. Example instructions are given below for a Postgres 15.0 database running on Ubuntu 22.04. For more information, or for alternative versions/operating systems, refer to the wal2json documentation

   • Update and upgrade apt if necessary.

     sudo apt update
     sudo apt upgrade -y

   • Prepare by making prerequisite installations.

     sudo apt install curl ca-certificates
     sudo install -d /usr/share/postgresql-common/pgdg

   • Import the repository keys for the Postgres Apt repository

     sudo curl -o /usr/share/postgresql-common/pgdg/apt.postgresql.org.asc --fail https://www.postgresql.org/media/keys/ACCC4CF8.asc

   • Create the pgdg.list file.

     sudo sh -c 'echo "deb [signed-by=/usr/share/postgresql-common/pgdg/apt.postgresql.org.asc] https://apt.postgresql.org/pub/repos/apt bookworm-pgdg main" > /etc/apt/sources.list.d/pgdg.list'

   • Use the Postgres Apt repository to install wal2json

     sudo apt update
     sudo apt-get install postgresql-server-dev-15
     export PATH=/usr/lib/postgresql/15/bin:$PATH
     sudo apt-get install postgresql-15-wal2json

4. Configure your database with wal2json enabled.

   • Edit your postgresql.conf configuration file so the following parameters are appropriately set.

     wal_level = logical
     max_replication_slots = 10
     max_wal_senders = 10
     

   • Restart PostgresSQL

   • Create a replication slot for tap-postgres.

     SELECT * FROM pg_create_logical_replication_slot('tappostgres', 'wal2json');

     You can also create the replication slot with any name you want, and set the replication_slot_name setting to that value.

5. Ensure your configuration for tap-postgres specifies host, port, user, password, and database manually, without relying on an sqlalchemy url.

6. Use the following metadata modification in your meltano.yml for the streams you wish to have as log-based. Note that during log-based replication, we do not support any replication key other than _sdc_lsn.

   metadata:
     "*":
       replication_method: LOG_BASED
       replication_key: _sdc_lsn