Our docs cover some of the data node management [0] and native replication [1] particular in more detail.
Right now you can use Postgres physical replication for the access node, with one as the primary (and handling DDL/writes), while Access Node replicas can scale reads to the rest of the cluster. As alluded to in the blog post, we'll probably look to enable clients to write directly to data nodes as a form of further scaling writes (so that the AN "primary" would only be strictly required for DDL), as opposed to a multi-master approach.
Right now you can use Postgres physical replication for the access node, with one as the primary (and handling DDL/writes), while Access Node replicas can scale reads to the rest of the cluster. As alluded to in the blog post, we'll probably look to enable clients to write directly to data nodes as a form of further scaling writes (so that the AN "primary" would only be strictly required for DDL), as opposed to a multi-master approach.
[0] https://docs.timescale.com/v2.0/using-timescaledb/distribute...
[1] https://docs.timescale.com/v2.0/using-timescaledb/distribute...