Linux: DB (LGWR) 10.2.0.4.0 hangs in io_getevents, when access to mirrored redo logs is lost
Hello,
normally, when Online Redo Logs are mirrored and an error occurs on one member of the redo log group, the member becomes STALE/INVALID
and the database continues running.
When you have some hardware problem on a storage system which results in re-mounting of the filesystems with ro on Linux
I recognized, that on Linux this behaviour (which is documented by Oracle) may be not valid - the database (LGWR) will hang in io_getevents.
After 10 minutes you will see WARNING:io_getevents timed out 600 sec in alert_${ORACLE_SID}.log.
You can only bounce the database with shutdown abort and LGWR will still be there (State: D (disk sleep)) ;-)
normally, when Online Redo Logs are mirrored and an error occurs on one member of the redo log group, the member becomes STALE/INVALID
and the database continues running.
When you have some hardware problem on a storage system which results in re-mounting of the filesystems with ro on Linux
I recognized, that on Linux this behaviour (which is documented by Oracle) may be not valid - the database (LGWR) will hang in io_getevents.
After 10 minutes you will see WARNING:io_getevents timed out 600 sec in alert_${ORACLE_SID}.log.
You can only bounce the database with shutdown abort and LGWR will still be there (State: D (disk sleep)) ;-)
0