Users executing complex CREATE TABLE AS SELECT (CTAS) statements in Greenplum 7.x may encounter a cluster-wide segment panic. The database logs typically show:
Segment log may show a stack trace like:
2026-07-03 10:45:33.036605 CEST,,,p655994,th0,,,2026-07-03 10:45:10 CEST,0,con107135,cmd176,seg15,,,,,"PANIC","XX000","Unexpected internal error: Segment process received signal SIGSEGV",,,,,,,0,,,,"1 0x7f071575b990 libpthread.so.0 <symbol not found> + 0x1575b990
2 0x99357c postgres slot_getsomeattrs_int (discriminator 5)
3 0x97dfe7 postgres <symbol not found> (execExprInterp.c:452)
4 0x9bc217 postgres <symbol not found> (palloc.h:194)
5 0x98c2c7 postgres <symbol not found> (execProcnode.c:646)
6 0x982f6b postgres <symbol not found> (execMain.c:2680)
7 0x983907 postgres standard_ExecutorRun (execMain.c:978)
8 0x983b05 postgres ExecutorRun (execMain.c:793)
9 0xba4c52 postgres <symbol not found> (pquery.c:252)
10 0xba517b postgres <symbol not found> (pquery.c:1525)
11 0xba6330 postgres PortalRun (pquery.c:1022)
12 0xb9f0e6 postgres <symbol not found> (postgres.c:1469)
"
This usually occurs on queries involving:
Greenplum: 7.x
Greenplum 7.0.0 introduced a new feature to push down "JOIN" below "UNION ALL" which can improve performance in some cases. This is disabled by default by the GUC, optimizer_enable_push_join_below_union_all.
When the GUC is enabled, the ORCA planner can created incorrect plans.
Disable the GUC:
# Change the setting in the config files:
gpconfig -c optimizer_enable_push_join_below_union_all -v off
# Reload the configuration
gpstop -u
gpconfig -s optimizer_enable_push_join_below_union_all -v off
As of July 2026, a fix is being developed and will be released in a future maintenance release.
Jira: TGP-27690