Users are not able to run queries in the database. Some utility commands (gpconfig with -s or -v flags) are failing.
Below is an example of the error that will be seen:
ERROR: Interconnect Error: Could not parse remote listeneraddress: 'mdw.example.local' port '52029': Name or service not known (seg0 slice1 172.1.1.1:20000 pid=41111) DETAIL: getaddrinfo() unable to parse address: 'mdw.example.local' SQL state: 58M01
This can occur for multiple reasons. One possible reason is that the customer has mistakenly configured the GUC "log_hostname" cluster-wide rather than on the master only.
Since 'gpconfig -s' and 'gpconfig -v' commands are not functional. Verify this by checking the gpconfig logs under /home/gpadmin/gpAdminLogs, or check the postgresql.conf under one of the segment data directories to confirm if 'log_hostname' was set to 'on'
If 'log_hostname' was mistakenly set cluster-wide, perform the following steps:
-Comment the line from all segment's postgresql.conf. This should not encounter the same issues as the '-v' and '-s' arguments.
gpconfig -r log_hostname
gpconfig -r log_hostname
-Reload the segment configurations
gpstop -u
-Confirm that the database is operational and queries are successful.
If you need to configure log_hostname, below is the gpconfig command that should be used:
gpconfig -c log_hostname -v on --masteronly
If there are questions regarding which GUCs should be set on the master only versus cluster-wide, please see the GPDB configuration parameters documentation:
https://gpdb.docs.pivotal.io/5280/ref_guide/config_params/guc-list.html