A gprecoverseg can fail with the following error message: gpadmin-[ERROR]:-Unable to parse recovery error. hostname: ## segments our of ## out of sync
The type of recovery does not matter in this case, as it can occur with all types of gprecoverseg.
This issue will happen even when all external indicators show a gprecoverseg should be successful. (E.G. hosts are active, system is up, full network connectivity, etc.)
Greenplum Database
If the system is unable to parse the error, it is possible this is because there is not enough space to write it.
The startup and recoverseg logs will show the recovery is successful up until it tries to write the conf file, and there the process will fail.
Check df -h, which should show the system to completely full on at least one host. Have the user try to drop as many databases as possible before attempting a full recovery. This will hopefully clear enough space on the mirrors to allow for the recovery to run again.
If that fails, the next step is for the customer to manually delete as much other data as possible before attempting the gprecoverseg again.