Operational issues when max_replication_slots is exhausted

On Mon, 2025-12-15 at 12:58 +0100, Ahmed Et-tanany wrote:

Our problem is that when our customers use up all available replication slots for logical replication,
our database management tasks that also require a slot fail (for example, creating the required
replication slot for a new physical standby). Since increasing `max_replication_slots` requires
a restart, we would like to avoid that if possible.

One idea we have considered is patching PostgreSQL to add a new GUC parameter that would allow
a superuser to reserve a certain number of replication slots usable only for management tasks.

Is this a known issue that might be addressed in PostgreSQL at some point? If not,
what would be a good way to solve this problem?

It is conceivable that somebody might change the behavior at some point (compare
"reserved_connections"). If you write or sponsor a patch, that would increase
the likelihood.

Right now, my only suggestion is to set "max_replication_slots" high.

Yours,
Laurenz Albe

On 12/15/25 03:58, Ahmed Et-tanany wrote:

Hello PostgreSQL community,

We have an issue related to `max_replication_slots` that I am not sure
would qualify as a bug,
so I thought I would ask here first.

Our problem is that when our customers use up all available replication
slots for logical replication,

The above would be the issue. Allowing customers to be DBA's for an
installation that they are not ultimately responsible for does not seem
sound to me. I would think a request mechanism is in order.

Thanks in advance.

Best regards,

--
Ahmed Et-tanany
Aiven: https://aiven.io/ <https://aiven.io/&gt;

--
Adrian Klaver
adrian.klaver@aklaver.com