This can happen if a DBMS server has been (accidentally) killed and 
	the csreport still shows the server's process_id "in use". The
	archiver won't start and the recovery is frozen.

	Check that the recovery has frozen with two successive logstats.
	Compare the log read i/o figures. Also have a look at lockstat.
	Check that there is no activity on the log device (eg with mon).

	If there is no activity:

		Kill the shutdown process.
		Kill the recovery process with a kill -HUP. The archiver 
		should also die.

		Retry the shutdown. It should now succeed.
		Clean up shared memory and semaphores (ipcs/ipcrm).
		Remove all entries from $II_SYSTEM/ingres/files/memory.
		Restart ingres and monitor II_ACP.LOG to see if there
		are any bad transactions. Check for inconsistent databases
		when the log is emptied. Recover databases if necessary.

		Consider applying a checkpoint.	

Ingres Q & A

To William's Home Page

© William Yuan 2000

Email William