Standard Postgres replication

TimescaleDB and PostgreSQL
1

TimescaleDB states that multi-node is deprecated from 2.13 onwards. I’d like to know what exactly this means. Will the standard Postgres replication continue to work? Will it also replicate databases that use the timescale extension? It’s not quite clear what the deprecation notice means exactly. I want to set up a HA database but if timescaledb somehow ‘breaks’ that, that would be a big problem.

timescale also has (or ‘had’) its own multi-node bits like ditributed tables. Is that the only thing that will no longer be developed after 2.13?

2

Hello @fredl,

Yes, that’s the only thing that is deprecated.

1 Like
3

TimescaleDB states that multi-node is deprecated from 2.13 onwards. I’d like to know what exactly this means. Will the standard Postgres replication continue to work?

I would expect the postgres built in streaming replication to work regardless as it does not really “know” if the data is timescale or not.

Will it also replicate databases that use the timescale extension?

I would expect it to.

1 Like
4

We had a concept named distributed hypertables which allows to create a multi-node server setup. It was also capable of using replicas as Timescale features are compatible with Postgresql, in general it does not change the behavior.

5

Hi @jonatasdp - could this be clarified where we have an on-prem deployment and would like to know if native replication from Postgres would work and similarly either an active-active or active-standby solution with postgresql exists.

6

There are two completely separate features:

  • Timescale offered until 2.13 that single hyper-table could be partitioned across multiple physical or virtual machines, on each of machine separate Timescale was installed. Such a table was called “distributed hyper-table”. This feature is discontinued.
  • PostgreSQL replication (which is independent from TimescaleDB) with primary (read+write) and hot-standby (read-only) and it works fine with Timescaledb. I have set this on-premis. On stand-by replica in postgresql.conf you have to have the same software installed as on primary and in postgresql.conf set shared_preload_libraries = 'timescaledb' just like on primary machine. All settings inside database (you know what ever you do after “psql” command is executed) must be set on primary machine. For example CREATE EXTENSION IF NOT EXISTS timescaledb; must be set on primary and this setting is replicated automatically to hot-standby replica. On hot-standby you can’t directly to database apply any settings or data changes, it has to be replicated

I’ve seen the following article pretty useful: How to Set Up Streaming Replication to Keep Your PostgreSQL Database Performant and Up-to-Date