Postgres FM · Hands-On Engineering Podcasts · November 2026
Postgres 19 removed a few configuration parameters, so a config file that still sets them stops the server from starting. Nikolay points out that the pre-upgrade check doesn't catch this, and recommends rehearsing upgrades on a full clone.
…feature, Nick, the self-tuning feature of the IO workers that I think Thomas Bonjo was working on. That's had a parameter removed. I haven't checked if that's one of them, but yeah, basically, there's some parameters that are no longer in Postgres that if you specify them, it won't start.
And be careful because you may have that set on, for example, a specific table. So one of the great features of Postgres is you can override things on various objects. And if you use that stuff on a specific object, you might not find it as fast. And so, yeah, so that's a red flag.
And PG upgrade check will report success. And then you just cannot start after upgrade. Yeah, that just means that, as usual, you need to test it thoroughly before going to production. For example, upgrading on a clone separately, checking everything works, and then already then proceeding to production, right? Testing on clones remains important. By the way, how we didn't touch any logical replication improvements, specifically replication of sequences.
Sequences, yeah.
The only thing left among big gaps is just DDL, right? The
only thing, that small thing. Yeah.