This is slightly different to , but was introduced by the same underlying change.
I can see multiple ways of solving this:
We can make the JDBC monitor/collector not use to connection pool and therefore free up the connections as fast as possible. This would increase the load on the target as every cycle will spawn a new connection
We can add another data-source which connects to the postgres database for our default monitoring config. This will multiply the pool resources even more.
We can make the opennms-admin data-source use the opennms database but then we have to tell people to create the database manually before installing OpenNMS
We can make the pool size configurable by data-source and add a new one for our default JDBC poller. This is .
PostgreSQL complains when creating a database from source database
template1with errorThe 48 sessions corresponds to the connection min pool setting in etc/opennms-datasources.xml which is set to the following
Checking against the pg_stat_activity table, it shows exactly 48 idle connections.
The behavior is the same as described in . I suspect it’s not fully fixed.