How to deploy modules separately in different servers without using a main configuration file?

[mrnadaara]

Apologies if I understood things incorrectly.

I'm writing some terraform logic that deploys a VPS and then runs a cloud-init config to setup the necessary dependencies. I would like to extend this behaviour by setting up PIGSTY modules in their respective VPS'.

The main server will have the infra module alongside other services that I've already deployed. A secondary server will be deployed that installs PGSQL as a primary. I would then add any extra servers needed for the PGSQL cluster as replicas. Each deployment has its own pigsty.yml. Whenever a new replica is added, the primary discovers it.

I'm unsure if this is possible as the documentation and examples I've read has all the configs in a single file. Whenever you want to add a new node, is the setup/installation of the pigsty environment handled by the admin node? My issue with this is that the automation of it would be cumbersome - create the VPS first, then add the PIGSTY SSH public key to the new server, then modify the pigsty.yml in the admin node and redeploy it. You can alternatively use the pg-add command but I'd much prefer the infrastructure changes are recorded in code.

My other concern is backups. PIGSTY recommends S/MNMD. MinIO is used but you can use an external S3 storage as backup. In the documentation for the S3 it says If you only have one node, a meaningful backup strategy would be to use cloud provider object storage services like AWS S3, Alibaba Cloud OSS, or Google Cloud, etc. To do this, you can define a new repository: Could you clarify if this limited to a single node setup or can this be used for a PGSQL cluster spanning multiple nodes?

Thanks

[Vonng]

1. You do need one admin node to run Ansible against the rest, but it can be anything that has SSH access to the other VPS — the first VPS itself, a separate VM, or even your laptop. It's just where ansible-playbook runs from.
2. External S3 works for multi-node clusters too. The docs phrase it around the single-node case because that's when people most often skip MinIO, but a full PGSQL cluster can use S3/OSS/GCS the same way — the cluster just shares one backup repo.

[mrnadaara]

Hey Feng, thanks for the reply.

If that's the case, is it possible to have the playbooks executed in a CI/CD environment? Is it possible to just extract the playbooks from the official script without the extra configuration steps it provides?