When attempting to upgrade Greenplum from 6.x to 7.x (e.g., 6.33.1 to 7.8.1) using the gpupgrade utility, the gpupgrade initialize command fails during the pg_upgrade check phase.
The CLI outputs generic RPC errors and exit statuses similar to the following:
2026/06/16 17:22:46 gpupgrade:gpadmin:<HOSTNAME>:224292 [INFO]: Running pg_upgrade checks... [IN PROGRESS]
2026/06/16 17:22:58 gpupgrade:gpadmin:<HOSTNAME>:224292 [INFO]: Running pg_upgrade checks... [FAILED]
2026/06/16 17:22:58 gpupgrade:gpadmin:<HOSTNAME>:224292 [INFO]: Initialize [14s]
2026/06/16 17:22:58 gpupgrade:gpadmin:<HOSTNAME>:224292 [INFO]: rpc error: code = Unknown desc = substep "check_upgrade": 3 errors occurred:
* check segment on host <HOSTNAME>: rpc error: code = Unknown desc = 4 errors occurred:
* check primary on host <HOSTNAME> with content 1: exit status 1
* check primary on host <HOSTNAME> with content 0: exit status 1
* check primary on host <HOSTNAME> with content 2: exit status 1
* check primary on host <HOSTNAME> with content 3: exit status 1
* check segment on host <HOSTNAME>: rpc error: code = Unknown desc = 4 errors occurred:
* check primary on host <HOSTNAME> with content 6: exit status 1
* check primary on host <HOSTNAME> with content 7: exit status 1
* check primary on host <HOSTNAME> with content 5: exit status 1
* check primary on host <HOSTNAME> with content 4: exit status 1
* check segment on host <HOSTNAME>: rpc error: code = Unknown desc = check primary on host <HOSTNAME> with content -1: exit status 1Source: Greenplum Database 6.x
Target: Greenplum Database 7.x
Tool: gpupgrade
Greenplum 7 removes support for several features and object types that were permitted in Greenplum 6. If your Greenplum 6 environment contains any of these deprecated objects, the pg_upgrade pre-checks will fail, halting the upgrade.
For example:
Tables compressed using QuickLZ.
Tables containing columns with the "unknown" data type.
For all items that pg_upgrade will check, please refer to: https://techdocs.broadcom.com/us/en/vmware-tanzu/data-solutions/tanzu-greenplum-upgrade/2-0/gp-upgrade/gpupgrade_initialize_checks.html
To resolve this issue, you must identify the unsupported objects, remove or modify them in the Greenplum 6 source cluster, and retry the initialization.
Step 1: Identify the Blocking Objects Check the master node at the following path to see which unsupported objects triggered the failure: ~gpadmin/gpAdminLogs/gpupgrade/pg_upgrade_<timestamp>/p-1/
If unsupported objects exist, pg_upgrade will generate .txt files containing lists of the specific database objects causing the failure. Look for files such as:
*_with_quicklz.txt
tables_using_unknown.txt
plpython2_dependent_functions.txt
Step 2: Review Detailed Fatal Errors (Optional but Recommended) For a detailed explanation of exactly why the failure occurred, review the internal pg_upgrade log on master node at: ~gpadmin/gpAdminLogs/gpupgrade/pg_upgrade_<timestamp>/p-1/log/pg_upgrade_internal.log
Example log output for an "unknown" data type blocker:
Checking for invalid "unknown" user columns fatal
| Your installation contains the "unknown" data type in user tables. This
| data type is no longer allowed in tables, so this cluster cannot currently
| be upgraded. You can remove the problem tables and restart the upgrade.
| A list of the problem columns is in the file:
| /home/gpadmin/gpAdminLogs/gpupgrade/pg_upgrade_20260617T085236/p-1/tables_using_unknown.txt
Step 3: Drop or Modify the Objects in Greenplum 6 Using the generated .txt files as a guide, log into your Greenplum 6 source database and either drop or modify the problematic objects so they comply with Greenplum 7 standards.
Step 4: Retry the Upgrade Once all conflicting objects have been cleared from the source database, run the initialization command again