Kbase P162757: Database does not complete crash recovery because all ai files become full backing out a transaction
| Autor |
Progress Software Corporation - Progress |
| Acesso |
Público |
| Publicação |
30/03/2010 |
|
Status: Unverified
SYMPTOM(s):
Database does not complete crash recovery because all ai extents are full.
AI extents switching during crash recovery
Variable database bi extent is several gigabytes in size.
(3773) Can't switch to after-image extent c:\path\dbname.a# it is full.
(5350) Database Server shutting down as a result of after-image extent switch failure.
FACT(s) (Environment):
Large files are enabled
All Supported Operating Systems
OpenEdge 10.x
OpenEdge Replication
CAUSE:
User process died while executing a long running open transaction that spanned multiple ai extent switches. The total switches triggered by this user are more than the number of ai files in the database structure. While going through transaction backout, the database was stopped and restarted. This requires not only undoing the original transaction but also the failed transaction backout. All of this causes the ai files to be written to.
FIX:
Copy the full ai files and mark them as empty.
If this is a replication enabled database and you are running 10.1B03 an higher, add enough additional ai files for the transaction backout to complete. Once crash recovery finishes and the source and target synchronize, the ai files can be backed up and marked as empty.
If this is a replication enabled database and you are running 10.1B02 and earlier, disable replication:
proutil dbname -C disablesitereplication
Backup the ai files, mark as empty, and begin process to re-source your OE Replication source and target databases.