The file config/manager/default-monitoring.yaml contains a metric called pg_replication_slots that at the moment can only be run if executed on the primary due to the execution of functions that cannot be executed on replicas. This is similar to the problem resolved by the patch contained in #1666 (issue #1604).

When running this query on a standby it returns WAL control functions cannot be executed during recovery:

 SELECT slot_name, slot_type, database, active, pg_catalog.pg_wal_lsn_diff(pg_catalog.pg_current_wal_lsn(), restart_lsn) FROM pg_catalog.pg_replication_slots WHERE NOT temporary 

This is also still present in the prometheus exporter: prometheus-community/postgres_exporter#547