Patroni: Automatic recovery from leader change in primary or secondary

• Whenever a new Leader is elected, the PgBouncer instance must be reconfigured to point to the new Leader. (true for both Primary and Secondary --> Distribution plus Geo concern)
  • Potential solution: HAproxy? see https://patroni.readthedocs.io/en/latest/README.html?highlight=haproxy#running-and-configuring
  • Always point to HAproxy and then this becomes an internal concern
  • We don't package HAproxy yet, integrate with Patroni
  • Issue needs to be created
• Whenever a new Leader is elected on the primary node, the Standby Leader on the secondary needs to be reconfigured to point to the new Leader.
  • A secondary can't connect to pgbouncer to support replication
  • HAproxy is needed on the primary for the secondary to replicate
  • Today we go directly to the node and you have to repoint manually
  • Not everything needs to be replicated again because of pg_rewind support
  • Issue needs to be created
  • Docs update gitlab#231252 (closed)