We were migrating a non-cdb 19.7 database from Solaris x86 to 19.14 pdb on Linux. datapatch threw the compilation error while executing prvthadoop.plb:<\/p>\n
Validating logfiles...done\nPatch 33515361 apply (pdb DB11): WITH ERRORS\n logfile: \/u00\/oracle\/orabase\/cfgtoollogs\/sqlpatch\/33515361\/24589353\/33515361_apply_DB1C_DB11_2022Jan24_12_28_40.log (errors)\n -> Error at line 274912: script rdbms\/admin\/prvthadoop.plb\n - Warning: Package Body created with compilation errors.<\/code><\/pre>\nprvthadoop.plb is a wrapper for prvthadoop1:<\/p>\n
:execfile := 'prvthadoop1.plb';<\/code><\/pre>\nprvthadoop1 creates the DBMS_HADOOP_INTERNAL package:<\/p>\n
CREATE OR REPLACE PACKAGE BODY DBMS_HADOOP_INTERNAL wrapped<\/code><\/pre>\nThe compilation of DBMS_HADOOP_INTERNAL failed because a table or view was missing:<\/p>\n
select name, text from dba_errors\n\nNAME\n--------------------------------------------------------------------------------\nTEXT\n--------------------------------------------------------------------------------\nDBMS_HADOOP_INTERNAL\nPL\/SQL: ORA-00942: table or view does not exist<\/code><\/pre>\ndba_errors doesn’t show which tables or views are missing, just the program line. But this information is to no avail because the package body is wrapped.<\/p>\n
I queried dba_dependencies on a healthy database to see which objects are referenced in DBMS_HADOOP_INTERNAL:<\/p>\n
select referenced_name from dba_dependencies where name = 'DBMS_HADOOP_INTERNAL'\n\nREFERENCED_NAME\n--------------------------------------------------------------------------------\n...\nDBA_HIVE_TABLES\nDBA_HIVE_COLUMNS\nDBA_HIVE_TAB_PARTITIONS\nDBA_HIVE_PART_KEY_COLUMNS\n...<\/code><\/pre>\nThe referenced DBA_HIVE* views exist on Linux:<\/p>\n
SQL> select object_name from dba_objects where object_name like 'DBA_HIVE%' ;\n\nOBJECT_NAME\n--------------------------------------------------------------------------------\nDBA_HIVE_TABLES\nDBA_HIVE_TABLES\nDBA_HIVE_COLUMNS\nDBA_HIVE_COLUMNS\nDBA_HIVE_DATABASES\nDBA_HIVE_DATABASES\nDBA_HIVE_TAB_PARTITIONS\nDBA_HIVE_TAB_PARTITIONS\nDBA_HIVE_PART_KEY_COLUMNS\nDBA_HIVE_PART_KEY_COLUMNS<\/code><\/pre>\nbut are missing on Solaris x86:<\/p>\n
select object_name from dba_objects where object_name like 'DBA_HIVE%' ;\n\nno rows selected<\/code><\/pre>\nThese views are created by the script cathive1:<\/p>\n
grep -il DBA_HIVE *\n...\ncathive1.sql\ndbmshadp1.sql\n...<\/code><\/pre>\ncathive1.sql is called by cathive.sql only on Linux (platform_id=13) and Solaris SPARC (platform_id=2), but not on Solaris x86 (platform_id=20):<\/p>\n
IF (:pfid = 13 OR :pfid = 2) THEN\n :execfile := 'cathive1.sql';\nELSE\n :execfile := 'nothing.sql';\nEND IF;<\/code><\/pre>\nselect platform_id, platform_name from v$transportable_platform where platform_id in (13,2,20);\n\nPLATFORM_ID\n-----------\nPLATFORM_NAME\n--------------------------------------------------------------------------------\n 2\nSolaris[tm] OE (64-bit)\n\n 13\nLinux x86 64-bit\n\n 20\nSolaris Operating System (x86-64)<\/code><\/pre>\nSimply put, datapatch on Linux expect Hadoop objects, but they won’t be there if the database was migrated from another platform.<\/b><\/p>\nWorkaround<\/h1>\n
You can first migrate to 19.13 PDB on Linux and than apply the 19.14 RU.<\/p>\n
I also tried the following unsupported way: created the Hadoop objects on Solaris x86 (in the lab) prior to migrating to 19.14 on Linux.<\/p>\n
@?\/rdbms\/admin\/dbmshadp1\n@?\/rdbms\/admin\/cathive1\n@?\/rdbms\/admin\/prvthadoop1.plb<\/code><\/pre>\ndatapatch ran without errors. You’d need to clarify with Oracle support before using it in the production.<\/p>\n
Update January 31, 2022<\/h1>\n
Oracle Support confirmed the workaround with creating the views on the source database.<\/p>\n","protected":false},"excerpt":{"rendered":"
datapatch is failing when migrating from Solaris to Linux Continue Reading