Thread

1. Proposal: Native High Availability and Automatic Failover in PostgreSQL

   Ale Rox <gitroxale@gmail.com> — 2025-06-25T07:26:00Z

   Dear all,
   
   I've been using PostgreSQL for a long time and consider it a solid and
   reliable database. However, I believe there is room for improvement in the
   area of high availability and automatic failover, inspired by features
   already available in MongoDB.
   
   Specifically, it would be extremely useful to have:
   
   - A built-in mechanism for automatic replication and failover, fully
   integrated into PostgreSQL, without relying on external tools like Patroni;
   - The ability to add new nodes to the cluster with automatic replication
   configuration;
   - Automatic failover behavior when a node goes down, allowing the cluster
   to continue operating, and resynchronizing the node automatically when it
   comes back online;
   - A JDBC driver (and ideally other drivers as well) that accepts a list of
   nodes in the connection string and can automatically detect the current
   primary node for write operations — similar to how MongoDB drivers handle
   replica sets.
   
   The goal would be to have a single, fully integrated solution for high
   availability, without needing to assemble multiple external components.
   This would make PostgreSQL significantly easier to manage and more
   competitive in environments where reliability and ease of use are key
   requirements.
   
   Are there any plans to introduce such functionality in the core PostgreSQL
   project in the future?
   
   Thank you for your attention and for the great work you do!
   
   Best regards,
   Rox Ale
   

2. Re: Proposal: Native High Availability and Automatic Failover in PostgreSQL

   wenhui qiu <qiuwenhuifx@gmail.com> — 2025-06-25T07:32:00Z

   HI Ale Rox
      I've been thinking like you for a long time, But it's not so easy to
   realise this, there are more oppositions than approvals，You can search for
   this title (Re: Built-in Raft replication)
   
   On Wed, Jun 25, 2025 at 3:26 PM Ale Rox <gitroxale@gmail.com> wrote:
   
   > Dear all,
   >
   > I've been using PostgreSQL for a long time and consider it a solid and
   > reliable database. However, I believe there is room for improvement in the
   > area of high availability and automatic failover, inspired by features
   > already available in MongoDB.
   >
   > Specifically, it would be extremely useful to have:
   >
   > - A built-in mechanism for automatic replication and failover, fully
   > integrated into PostgreSQL, without relying on external tools like Patroni;
   > - The ability to add new nodes to the cluster with automatic replication
   > configuration;
   > - Automatic failover behavior when a node goes down, allowing the cluster
   > to continue operating, and resynchronizing the node automatically when it
   > comes back online;
   > - A JDBC driver (and ideally other drivers as well) that accepts a list of
   > nodes in the connection string and can automatically detect the current
   > primary node for write operations — similar to how MongoDB drivers handle
   > replica sets.
   >
   > The goal would be to have a single, fully integrated solution for high
   > availability, without needing to assemble multiple external components.
   > This would make PostgreSQL significantly easier to manage and more
   > competitive in environments where reliability and ease of use are key
   > requirements.
   >
   > Are there any plans to introduce such functionality in the core PostgreSQL
   > project in the future?
   >
   > Thank you for your attention and for the great work you do!
   >
   > Best regards,
   > Rox Ale
   >
   

3. Re: Proposal: Native High Availability and Automatic Failover in PostgreSQL

   Greg Sabino Mullane <htamfids@gmail.com> — 2025-06-25T15:30:31Z

   On Wed, Jun 25, 2025 at 3:26 AM Ale Rox <gitroxale@gmail.com> wrote:
   
   > Specifically, it would be extremely useful to have:
   >
    (snip wishlist)
   
   >  Are there any plans to introduce such functionality in the core
   > PostgreSQL project in the future?
   
   
   Getting failover to work, and work CORRECTLY[1], is an extremely
   complicated task, fraught with tons of complicated edge cases and risks. It
   may arrive "in core" someday, but it's going to be a very long road. I
   would suggest starting on one of your bullets. Pick as small of a feature
   as you can, then expand on its use case and all the specific items it would
   need to do. Look at Patroni (current best-in-class failover system for
   Postgres) and see how it does it, then try to map how Postgres itself could
   do the same.
   
   [1] MongoDB does not have a good rep in this area
   
   Cheers,
   Greg
   
   --
   Crunchy Data - https://www.crunchydata.com
   Enterprise Postgres Software Products & Tech Support
   

4. Re: Proposal: Native High Availability and Automatic Failover in PostgreSQL

   Greg Sabino Mullane <htamfids@gmail.com> — 2025-06-25T15:33:16Z

   On Wed, Jun 25, 2025 at 3:26 AM Ale Rox <gitroxale@gmail.com> wrote:
   
   > - A JDBC driver (and ideally other drivers as well) that accepts a list of
   > nodes in the connection string and can automatically detect the current
   > primary node for write operations — similar to how MongoDB drivers handle
   > replica sets.
   >
   
   This one could be done with the help of one of my outstanding patches:
   
   https://www.postgresql.org/message-id/flat/CAKAnmm+T-CEDLmRezWfH-7ZEsFfD_kU2KY1TgB288K+wprB_4Q@mail.gmail.com
   
   
   -- 
   Cheers,
   Greg
   
   --
   Crunchy Data - https://www.crunchydata.com
   Enterprise Postgres Software Products & Tech Support