Applies to:
Oracle Server - Enterprise Edition - Version: 10.2.0.1 to 10.2.0.3
This problem can occur on any platform.
Symptoms
During normal database activities some ORA-600s are reported. They are:
ORA-00600: internal error code, arguments: [17147], [0xC0000001B67EED68], [], [], [], [], [], []
ORA-00600: internal error code, arguments: [KGHLKREM1], [0xC0000001F4CE11B8], [], [], [], []
ORA-7445: exception encountered: core dump [kghlkremf()+33] [SIGSEGV]
The sql that causes the error is not located in the trace file.
Call stack:
kghfrmrg kghfre qsmqktcc ktcdso ktcrcm ktdcmt k2lcom k2send xctctl xctcom_with_options opicom
or
k2send xctctl xctcom_with_options opicom opiodr
or
kghalo ktcccadd kwqidracbk kwqidcpmc kwqidafm0 kwqididqx kpoaqdq
or
kghadd_reserved_extent kghget_reserved_extent kghgex kghfnd
Cause
These internal errors are reported in several bugs closed as duplicate of unpublished Bug 5736850.
The bug involves abnormal termination of queries involving bloom filters.
This manifests itself as memory corruptions. The memory was freed, reallocated to some other client, and
corrupted but slaves were still writing into it.
Solution
This fix cannot be provided as a one-off because it changes fixed sga structures and introduces a
new wait event.
It can be provided only as part of a complete patchset so the first version that will include this
fix is 10.2.0.4.
Setting _bloom_filter_enabled = false is a possible workaround.
This parameter can be set dynamically:
SQL> connect / as sysdba
SQL> alter system set "_bloom_filter_enabled"=false scope=both;