I'm running the database in archivelog mode. The alert log has the following of and on...
LGWR: prodding the archiver
... but yesterday afternoon onwards, this kept repeating in the alert log and the server could not allocate new log as the online logs needed archiving...
Sorted this out, by acknowleging message on console and then giving ...
Code:
archive log all;
Trying to investigate why this happened ?? All applications using this database were affected! Everyone one was trouble-shooting the other tiers of the application but the database. Any pointers??
BTW, the following info was also in the event viewer...
Archive process error: ORACLE Instance blaze - Can not allocate log, archival required.
I suppose the problem was with the archive process not waking up inspite of being prodded by the log writer.
The error ...
LGWR: prodding the archiver
... comes continously in the alert log for more than 2000 times, one line for every second or so it seems! (the alert log file increased from 1.7 mb to 5.5 mb).
Maybe I need to also mention that this is the first time I'm getting this problem after configuring for archivelog mode 6 months back. The database has 3 log groups of 5MB each.
BTW, the following info was also in the event viewer...
Archive process error: ORACLE Instance blaze - Can not allocate log, archival required.
This is a typical error when all log files get full and no arhiving is enable..
1. Oracle tries to write new transactions to logfile
2. Old logs (which are already full) aren't got archived automatically
3. Doesn't able to allocate new log
If you read my post carefully, you will notice that the db is in archivelog mode...so obviously log_archive_start is true!
I repeat, that the link you posted speaks about a situation where a unix file system is full and the logs are diverted to another file system. I like to distinguish between 'logs being full' and 'disk being full'. Ofcouse, in my situation, 3 redo log groups are full and not archived because of the archiver not responding (it seems!) inspite of the efforts of the log writer to awaken them.
I hope I make myself clear.
Metalink, speaks about such errors ocurring because of OS or hardware problems. Not much similar situations where you have... LGWR: prodding the archiver ... repeating continously in the alert log.
Yes, it looks like there is a problem with ARC0 process. Raise a TAR with metalink. Bounce the database for temporary solutions. Before that do take a trace of all the error msgs.
How busy was your system? IE, if your process list was too long (connections, other jobs, other apps, etc),the archiver process may be too far down to start?
I am not 100% sure, but I believe I have run into this once or twice in Unix over the years.
Joseph R.P. Maloney, CSP,CDP,CCP
'The answer is 42'
Reply With Quote
Bookmarks