EDB Postgres Distributed 5.8.1 release notes v5.8.1

The --new-socketdir CLI option and corresponding PGSOCKETDIRNEW env variable was introduced to address a scenario where the old and new socket dirs are different. This fix addresses the issue by fetching the value of new socket dirs from the initialized new server.

This fixes the command crash if the CLI cannot find the binaries like postgres, pg_ctl on the expected path. By default it searches the binaries at the path of the pgd binary itself.

The mechanism for determining whether user can be granted permissions on BDR extension objects was failing for EPAS-specific objects. This is a regression from PGD release 5.7.0 and is now fixed.

Physical join using bdr_init_physical resulted in creation of an extra slot after the physical join is over, and can result in holding up WAL. This is a regression from BDR release 5.7.0. This is now fixed.

This release has fixed several issues related to node synchronization in BDR. The node sync process now properly cleans up slots and subscriptions when any of the source, target, or origin parts are involved. Additionally, the node part operation has been fixed to prevent hanging during node sync operations. Subscriber-only nodes will also now be correctly synced.

The check is renamed to be consistent with the terminology used at other places in PGD.

This fixed the command crash issue if CLI cannot fetch value of a GUC for some reason.

Rejoin of parted node with same name gets stuck in JOIN_START state. This problem was fixed.

Fixed non-transactional WAL message acknowledgment by downstream that could cause a WAL sender to never exit during fast database system shutdown.

DROP_REPLICATION_SLOT command for physical / non-existing slots was failing with an ERROR.