Enhancements, bug fixes, and other changes in EFM 5.4 include:
| Type | Description | Addresses |
|---|---|---|
| Enhancement | Failover Manager now supports retrieving database passwords through an external script. | 49606 |
| Enhancement | Failover Manager now includes efm commands for retrieving streaming replication information. | 54883 |
| Enhancement | Failover Manager now includes an HTTP endpoint for load balancers to locate the primary node. | |
| Enhancement | Failover Manager now uses the /run and /run/lock directories instead of the legacy /var/run and /var/lock symlink directories. | |
| Enhancement | Standby creation through the efm command now supports copying an external configuration directory (e.g. on Debian). | |
| Enhancement | During an agent shutdown, the shutdown code will now verify that the lock file has been removed before reporting that the stop is complete. | |
| Enhancement | Failover Manager agents now log timing information for external processes that last over a minute. | |
| Enhancement | Systemd will no longer report a failed state for agents that are stopped through the efm stop-cluster command. | |
| Enhancement | Failover Manager was upgraded to use Log4J version 2.25.5. | |
| Enhancement | Failover Manager was upgraded to use Jakarta Mail version 2.0.2. | |
| Enhancement | Failover Manager was upgraded to use PostgreSQL JDBC Driver version 42.7.12. | |
| Bug Fix | Fixed an edge case that could cause a false promotion in a gradual network split of an even number of nodes. | |
| Bug Fix | If the minimum.standbys property is used, having too few standbys will no longer block the manual switchover case. | |
| Bug Fix | Fixed a bug that lost standby priority in certain cases when priority.standbys was used and promotable was false for some standbys. | |
| Bug Fix | Fixed an edge case that occasionally prevented agents from joining into a single cluster if started simultaneously. | |
| Bug Fix | Fixed an edge case that could prevent a rebuilding agent from finding the current primary agent during a failover. | |
| Bug Fix | Fixed an edge case that could cause agents to use an incomplete member list when rejoining the cluster after efm reset-members is run. |