Alexander Scheel

Hey @klaus-sap -- one note as well I was thinking about: in supporting PostgreSQL, I'll likely be building a mechanism for detecting/waiting for standby nodes to hit certain WAL checkpoints....

@tdiesler Just since I stumbled across this in the context of [OpenBao](https://openbao.org/), note that this usually means you haven't set [an API address (`api_addr`) in your configuration file](https://openbao.org/docs/configuration/#high-availability-parameters). Per (MPL-licensed)...

@weberval I'm hesitant to remove the old UI until we have the new UI fully accepted. Past discussions around ember upgrades have shown blockers and we're unable to make progress...

Looks like test failures since: > Quay Container Registry is currently in read-only mode. Pulls and other read-only operations will succeed but all other operations are currently suspended.

@satoqz This should be updated :-)

@satoqz This could use a re-review, I pushed two new commits to guard it only on active, HA-enabled nodes.

@satoqz I've rebased this and checked that it doesn't conflict with #1986. I also dropped the change around active node, based on discussion here: https://github.com/openbao/openbao/pull/1986#discussion_r2524451674

@satoqz @genelet Just to reason through the implications of this: 1. It means we're free to write from standby nodes still, if we bypass the barrier. 2. Active node lock...

> ...unless we hit the fence? Are you describing the state before this PR or after? Or is this just a remark that the underlying postgres client is fundamentally never...

> but they still _have_ an _instance_ of the lock, no? So wouldn't `validateFence/IsActivelyHeld` fail? No, they don't. You don't want them to, either. You could call a two step:...