What was the whole point of this divergence :)
We use replication slots and it is for this reason we are able to scale our applications by redirecting reads and writes backed by haproxy and pgbouncer. with the health check of replication lag. It works awesome and at a much much lower cost to what aws used to charge us for. (1 master, 3 slaves .. that goes to 7-8 during peak days)
With the use of logical replication we were able to create a multimaster mesh that allowed the writes to certain tables across the ocean and all regions subscribed to it.
As a result, at any given time all 5 availability. Zones had same data and we would sustain an occasional WAN outage between two DCs.
We use around 300 Postgres servers, everything managed via puppet and foreman.
And I have not seen any single issue w r t the one reported for any of our setups.
So I guess it is generalising just based out of some threads :)
I do not use db2 but there has never been a need for one in my company as we have multiple design patterns of Postgres architecture based of requirements.
>
> You're not forced to use slots. Their purpose is to allow to force the
> primary to keep necessary resources around. Which also allows to get rid
> of the archive in some setups.
Thanks.
Disclaimer: We don't use replication as we piggy back on AWS HA.
The reason why I posted this is because majority of replication related messages in this forum
is about slots :-)
--
