PostgreSQL includes Patroni Cluster which facilitates the creation of a High Availability (HA) configuration. Bobcares, as a part of our PostgreSQL Support offers solutions to every query that comes our way.

PostgreSQL – Patroni Cluster

Patroni is a part of VMware Postgres. A High Availability (HA) setup can be easily created using the cluster manager program Patroni. The setup and upkeep of PostgreSQL HA (High Availability) clusters may be automated and customized with Patroni, cluster management. It also offers streaming replication and automatic database failover.

For TCP and HTTP-based applications, HAProxy enables load balancing and proxying.

The distributed key-value store, Etcd, offers a means to store data that a distributed system or cluster of workstations needs to access. To keep the PostgreSQL cluster working, use Etcd to store the cluster’s state.

The database server host IP address and port must be set to the HAproxy virtual IP address and port.

Three data nodes in a cluster are the minimal config advised by VMware. Also, three nodes are needed for the Postgres database, three for etcd, and one for a load balancer, for a total of seven nodes in the three-node arrangement. For maximum endurance, etcd should mirror the database cluster’s topology. In order to reduce the risk of a single point of failure (SPOF), the load balancer can be setup on multiple hosts.

The etcd topology

It is recommended that the network topology of the etcd nodes match the topology of the data nodes; for instance, three data nodes require three etcd nodes. Postgres and etcd should be on separate disks if that is not practicable.

The advantages of the VMware-recommended topology include:

Each data node would still be able to communicate with its own etcd node in the case of a network partition. Also, by doing this, a scenario where Postgres nodes are capable of replication but Patroni is unable to write to etcd prohibits Patroni from managing the instances.

Disk I/O throughput needs for etcd are very strict. It is conceivable for etcd and Postgres to starve each other for disk I/O if they are co-located on the same physical host, which could result in cascading failures.

The resilience of both etcd and Postgres is reduced at the same time when the data host must be start again (for example, for security upgrades). As a result, the system is more vulnerable during the restart.

[Looking for a solution to another query? We are just a click away.]

Conclusion

To sum up, our Support team went over the PostgreSQL Patroni Cluster details.