APPLIES TO:
Oracle GoldenGate - Version 10.4.0.0 and later
Information in this document applies to any platform.
SYMPTOMS
- Oracle GoldenGate (OGG) extract abends with error
- Most likely Oracle RAC DB with log files on ASM with readahead enabled is in use
- Using parameter TRANLOGOPTIONS _NOREADAHEAD ANY doesn't help
CAUSE
The issue happens under rare conditions when the extract reads the log where one log write buffer contains a single record that happens to completely fill up the block it occupies, and the log cursor is at EOF.
This issue had been seen mostly on logs under ASM.
SOLUTION
Workaround:
Restart the extract. The extract will get restarted without any issue and will continue processing the records.If needed you could have the Autorestart parameter for the extract in the manager.
Solution:
Apply an OGG build which contains the fix for bug 12629456 + 13072229
The issue is fixed in OGG 11.2.1.0.0
The fix is included in OGG 11.1.1.1.2
The same issue was later reported in v11.1.1.1.5 via Bug 16865401. The issue happened here when the redo record size is greater than 65k.
The issue is fixed in 12.1.x and it is backported to 11.2.1.0.7 & 11.1.1.1.5_09 version.
Case 2
******
OGG v12.1.x Classic extract abended with
ERROR OGG-01028 Detect partial pdata at rba xxx without coinciding crash recovery marker record in log with seqno = xxxx.
This is due to bug 18643085.
The issue is fixed in 12.1.2.1.0, and workaround is to use the parameter "tranlogoptions _skipincompletelogdata"