>> if we use the --link option for pg_upgrade, what would be the rollback plan in case issues arise during or after the upgrade?
restoring a backup would be your option in such a case I guess.
On 15.06.25 12:05, Kushal Shri wrote:
Currently, our plan is to test pg_upgrade using the --link option to upgrade from PostgreSQL 9.5 to 12, then upgrade Django to the latest supported version, and finally perform another pg_upgrade to PostgreSQL 16.
Since Django 2.0 is officially tested and supported only up to PostgreSQL 10, would you recommend upgrading to PostgreSQL 10 using pg_upgrade as the first phase, followed by logical replication to PostgreSQL 16 as the final target after upgrading Django as well?
and, if we use the --link option for pg_upgrade, what would be the rollback plan in case issues arise during or after the upgrade?
I'm sorry if this part of the answer should not be given in this context, but I happen to be in this list (as a lurker for most of the time) and be a Django user.
Till Django 4.2 (LTS) postgres 12 is supported. Django 5.2 (LTS also) supports postgres 17+.
On Wed, 2025-06-11 at 16:39 +0545, Kushal Shri wrote:
> We are in the process of planning a PostgreSQL upgrade and would greatly
> appreciate any insights, best practices, or lessons learned from your experiences.
>
> Current Setup:
> - PostgreSQL version: 9.5
> - Django version: 2.0
> - Database size: ~3 TB
> - One Master 4 Replicas(Streaming Replication)
>
> We are taking a phased approach to this upgrade. In Phase 1, our goal is to
> upgrade PostgreSQL from v9.5 to v12, ensuring compatibility with our current
> Django 2.0 application.
That's already wrong. PostgreSQL v12 is out of support. Upgrade to v17 or v18.
> As part of this, I would like to understand:
>
> - What is the highest PostgreSQL version reliably supported with Django 2.0?
You'll have to ask Django what they support.
> - What upgrade strategies or methods have worked well for you in similar environments?
pg_upgrade --link
> - Any gotchas or performance considerations to keep in mind with large databases
> during such upgrades?
Make sure you don't have millions of Large Objects, otherwise the upgrade
will be very slow and memory consuming.
Yours,
Laurenz Albe
--
