Need guidance on partioning

Define large.

IMO, I don’t partition until 750M to 1B rows.

Maybe 500M is the table is heavily used or very wide.

Partitioning under those row counts without working knowledge can hinder performance.

From: srinivasan s <srinioracledba7(at)gmail(dot)com>
Sent: Wednesday, May 22, 2024 5:04 AM
To: pgsql-admin(at)lists(dot)postgresql(dot)org
Subject: [EXTERNAL] Need guidance on partioning

Hello everyone,

I hope you are all doing well.

I am seeking guidance on how to implement partitioning in PostgreSQL.

We have a large table that currently does not have any partitioning, and we have two requirements for removing old data from this table. We are looking to create a new table with partitioning.

1. The first requirement is to delete all records from the table that are older than 18 months. I believe we can achieve this by using range partitioning on the timestamp column.

2. The second requirement is to remove data from the table when a user leaves the organization. We have the account ID and user ID in the same table.

Could someone please offer guidance on selecting the appropriate partitioning method (range, sub-partition, or composite)?

Additionally, not all queries use the timestamp column in the WHERE condition. Is it mandatory to use the partition key in the WHERE condition to benefit from partitioning? Can we create a composite index that combines the partition key column with other columns used in the WHERE clause? Would this be beneficial?

Thank you.