Someone already mentioned the view permissions shift here which looks at caller permissions versus the view creator permissions, that will be compelling in a lot of cases so like this is something that I'd probably raise internally for my team for a few of the apps we maintain and then have a back and forth with the principals, if we like it then go through our current list of work with a PM and our manager and see if it makes sense to do now with the current pipeline of work etc.
As long as you stay well ahead of that, you are fine.
However, I'd strongly recommend testing your applications with new versions as they come out, or in beta (even better). You may learn about new features, find bugs in your application, find bugs in your migration processes, etc. If you bring any problems to the community right away, then by the time you actually want to upgrade, the problems may already be solved in the newer release.
I think that's a nice compromise; although not a "real" test as the load is a lot less. Also things may work in dev but not production, but unless you explicitly use new features that's rare.
For larger setups you can us canary servers and whatnot, but I prefer to work at small companies and keep my sanity.
I do like the idea of "canary versioning" some stuff. It's a bit of a game theory issue isn't it, since we all need prod to work at all times and can't risk a .0, and a major version difference between dev and prod can mask some nasty issues in my experience.
Unluckily, chances they will report bugs not that high - at least I don't expect bugs related to any moderate size/loaded databases be found by this way.
I use it in a container mainly, and MariaDB has an option to enable auto upgrades, but postgres requires you to manually upgrade every time.
For large databases where you care about uptime upgrading is usually a huge project anyway, since you have to make sure that nothing breaks due to the new version (check query plans, check if the new version changed anything, etc). You also probably want to use logical replication for the upgrade. There are things that PostgreSQL can improve of course.
As an aside I recently was involved in an upgrade from 9.6 where we used logical replication and pg_dump (yes, we really used pg_dump on a 20+ TB database). The reason for using pg_dump was because we have some tables which bloated very fast meaning we could not hold a snapshot open on the master long enough to do the initial data copy so the initial copy was done with pg_dump from a paused replica. And to not lose any data during pg_dump/pg_restore we kept a logical replication slot alive (but unused until pg_restore was complete). That was an interesting project, especially since Googling did not turn up anyone who had done the same.
The only change is that versions used to be X.Y.Z, and now it just used X.Y. Now, X gets incremented for every major version, so it seems like it's moving faster.
(It is not my will that these databases aren’t on newer versions, I would very much like them to be)
Basically I'm just curious what broke for you
I believe minor version upgrades are easily enough since they maintain disk format compatibility and you can simply spin up a new replica server under the new postgres version and then when it's synched with the main database you can make the switch and kill the old version. I'm not so sure about migrating to a new major version (ex: 9.x to 15.x).
Notably, Postgres has more formal support for logical replication. Logical replication is distinct from the normal replication in that it forwards actual SQL to the replica rather than binary write ahead logs.
The SQL is portable across versions more than WALs are. So with logical replication you can copy a live db to a new version. Then you just need some (ideally) brief down time to swap out the servers.
There are solutions for 9.x, but it got much better later on.
Very excited for this year, as we'll be using logical replication to migrate, which should make this much less risky. (since the old systems are untouched).
In the past, we used pg_upgrade for in place upgrades, and even though we took snapshots before, still seemed a bit stressfull when it happened.
We generally stay one major version behind on the main production clusters. The major reason is usually just scheduling - PG upgrades are large projects that touch a lot of things, take a lot of prep and impact every part of our business, so scheduling it is always a fraught negotiation.
If you don't have a lot of complexity, I recommend finding a cadence that doesn't hurt, but sticking with it. Once you get far enough behind, it becomes harder to upgrade, mostly because you also weren't upgrading surrounding tooling, so you end up changing a lot of things at once.
As your environment grows in complexity and use, at some point I think you'll find that nontechnical concerns start to dominate these decisions.
This may depend on your situation. For example, if you were running a large IT org with a few staff members with a lot of installed software, the cycle to process and check correct operation of each software under your management may be long enough that you'd start soon: by the time you are working on the last few systems under your responsibility, EOL for that Postgres version may be in sight.
You may also be working on a very new software project, where bugs are of lower consequence and you'd like to forestall the friction of an upgrade motivated by EOL. Or, you may find yourself in a lull in useful development and this is a good way to get ahead of maintenance, effectively reducing pressure in the future. In this case, I'd also consider upgrading...after a few months. Even Postgres has somewhat iffy .0 releases.
My own experience is in vertically integrated teams where we control all the software, so we typically upgrade every second or third version, leaving over a year to resolve issues before EOL of the version we're on.
We use AWS Aurora with Postgres and we autoscale our cluster. Because most upgrades have a few perf improvements we have found that we save a bit of money every time we upgrade, either in reduced CPU (which translates to running fewer instances) or in better memory efficiency (which boosts buffer_cache_hit_ratio and therefore reduces IO).
We use django and some other stuff that has very decent support for postgres, and in my experience postgres updates rarely have breaking changes that require much if any changes for us. This also keeps us from being "trapped" on an old version.
I reiterate from all my systems management - do not update to a .0 release of anything unless there is a critical security update and you're forced to.
EDIT: Oh directly for the question - eh depends. Every 2 to 3 max, and usually after it's tested by others more.
And new projects generally use whatever is newest unless it's a .0 release.
A very cool PostgreSQL feature is that, if you install a second instance on the same server, it will by default use a different path and port to run the second instance.
root@example:~# pg_lsclusters
Ver Cluster Port Status Owner Data directory Log file
11 main 5432 online postgres /var/lib/postgresql/11/main /var/log/postgresql/postgresql-11-main.log
14 main 5433 online postgres /var/lib/postgresql/14/main /var/log/postgresql/postgresql-14-main.log
I have only done this migration process on Debian systems but in my experience it worked fantastic!
I haven't worked on anything recently with a good performance testing system so that weighs in a bit, too (there's not a high degree of confidence the app won't hit a weird perf regression or edge case)
If it's using a managed service, that weighs in, too. Something like AWS RDS Postgres usually has a 6 month lag + give it some additional time to work out any AWS control plane "surprises"
Do admins usually upgrade their DBs with each major release? I'm guessing it's highly contextual and depends on how easy it is to do so, but I've heard about places that never upgrade until it's a huge problem for them to do so (in order to avoid an even worse problem.)