Hello,
I have a PostgreSQL HA configured using pg_auto_failover. I have 2 data nodes and one monitor node.
I have experienced a situation couple of times, when both the primary and secondary nodes are left in read/write. The primary node was unresponsive due to excessive memory utilization and paging. Then the monitor promoted the secondary node (which is expected) and seems like tried to bring down the primary. But the primary never went down. It became responsive after few minutes(when the memory came back to normal). This led to making both primary and secondary databases left in read/write mode.
Is this some kind of a bug or to be managed externally?
Hello, I have a PostgreSQL HA configured using pg_auto_failover. I have 2 data nodes and one monitor node. I have experienced a situation couple of times, when both the primary and secondary nodes are left in read/write. The primary node was unresponsive due to excessive memory utilization and paging. Then the monitor promoted the secondary node (which is expected) and seems like tried to bring down the primary. But the primary never went down. It became responsive after few minutes(when the memory came back to normal). This led to making both primary and secondary databases left in read/write mode. Is this some kind of a bug or to be managed externally?