A First Cluster Deployment v23
In this short tutorial, we are going to work through deploying a simple M1 architecture deployment onto a local Docker installation. By the end you will have four containers, one primary database, two replicas and a backup node, configured and ready for you to explore.
For this example, we will run TPA on an Ubuntu system, but the considerations are similar for most Linux systems.
If you're an EDB customer, you'll want to follow the EDB Repo instructions which will install the TPA packages straight from EDB's repositories.
If you are an open source user of TPA, there's instructions on how to build from the source which you can download from Github.com.
Follow those guides and then return here.
As we said, We are going to deploy the example deployment onto Docker and unless you already have Docker installed we'll need to set that up.
On Debian or Ubuntu, install Docker by running:
For other Linux distributions, consult the Docker Engine Install page.
You will want to add your user to the docker group with:
Currently, TPA requires Cgroups Version 1 be configured on your system,
and if you do not see a reference to
tmpfs in the output, you'll need to disable cgroups v2.
To make the appropriate changes, then update Grub and reboot your system with:
Giving a user the ability to speak to the Docker daemon lets them trivially gain root on the Docker host. Only trusted users should have access to the Docker daemon.
The next step in this process is to create a configuration. TPA does most of the work for you through its
configure command. All you have to do is supply command line flags and options to select, in broad terms, what you want to deploy. Here's our
tpaexec configure command:
This creates a configuration called
demo which has the M1 architecture. It will therefore have a primary, replica and backup node.
--platform docker tells TPA that this configuration should be created on a local Docker instance; it will provision all the containers and OS requirements. Other platforms include AWS, which does the same with Amazon Web Services and Bare, which skips to operating system provisioning and goes straight to installing software on already configured Linux hosts.
--postgresql 15, we instruct TPA to use Community Postgres, version 15. There are several options here in terms of selecting software, but this is the most straightforward default for open-source users.
--enable-repmgr tells TPA to use configure the deployment to use Replication Manager to hand replication and failover.
--no-git turns off the feature in TPA which allows you to revision control your configuration through git.
Run this command, and apparently, nothing will happen on the command line. But you will find a directory called
demo has been created containing some files including a
config.yml file which is a blueprint for our new deployment.
Now we are ready to create the containers (or virtual machines) on which we will run our new deployment. This can be achieved with the
provision command. Run:
You will see TPA work through the various operations needed to prepare for deployment of your configuration.
Once provisioned, you can move on to deployment. This installs, if needed, operating systems and system packages. It then installs the requested Postgres architecture and performs all the needed configuration.
You will see TPA work through the various operations needed to deploy your configuration.
You can quickly test your newly deployed configuration using the tpaexec
test command which will run pgbench on your new database.
To get to a psql prompt, the simplest route is to log into one of the containers (or VMs or host depending on configuration) using docker or SSH. Run
to ping all the connectable hosts in the deployment: You will get output that looks something like:
Select one of the nodes which responded with
SUCCESS. We shall use
uptake for this example.
If you are only planning on using docker, use the command
docker exec -it uptake /bin/bash, substituting in the appropriate hostname.
Another option, that works with all types of TPA deployment is to use SSH. To do that, first change current directory to the created configuration directory.
For example, our configuration is called demo, so we go to that directory. In there, we run
ssh -F ssh_config ourhostname to connect.
In both cases, you will be logged in as a root user on the container.
We can now change user to the
postgres user using
sudo -iu postgres. As
postgres we can run
psql. TPA has already configured that user with a
.pgpass file so there's no need to present a password.
And we are connected to our database.
You can connect from the host system without SSHing into one of the containers. Obtain the IP address of the host you want to connect to from the
We are going to connect to uptake, so the IP address is 172.17.0.11.
You will also need to retrieve the password for the postgres user too. Run
tpaexec show-password demo postgres to get the stored password from the system.
Assuming you have a Postgresql client installed, you can then run:
Enter the password you previously retrieved.
You are now connected from the Docker host to Postgres running in one of the TPA deployed Docker containers.