Database performance considerations for multi-instance deployment #366
Replies: 3 comments
-
The main performance issue is fixed but the note is remained in docs. Any other usages are welcome for high availability, but I want to leave it to users. As far as I know, most users has only ability to use docker, let alone k8s. So I didn't put much thing about how to use k8s, just a demo with tiny explanations there. I think it's not my duty to teach people about cloud native stuff. However, if you want to help me make the docs better, that's definitely welcomed. |
Beta Was this translation helpful? Give feedback.
-
if you want to contribute to the docs, please make pr to github.com/gzctf/website And you are welcome to join the discord and telegram group if you want. |
Beta Was this translation helpful? Give feedback.
-
Yes, such PR is definitely welcomed. |
Beta Was this translation helpful? Give feedback.
-
https://gzctf.gzti.me/guide/issues/qa.html#anything-worth-noting-when-deploying-to-a-multi-node-cluster
The experienced performance issues might be because of the overhead of creating/closing connections. This is due to the lack of using a "connection pooler". Its job is to minimize this overhead.
It is recommended to use PgPool or PgBounder. So instead of deploying a simple raw YAML manifest to deploy postgres. A sophisticated HA setup should be deployed. For example postgresql-ha helm chart created by bitnami.
I have been recently working on a helm chart that instantly deploys GZCTF with redis-ha (ArgoCD uses this) and postgresql-ha as chart dependencies with a single command. If such contributions are welcome let me know I can submit a PR
Beta Was this translation helpful? Give feedback.
All reactions