Oracle 10g lockup: heap exceeds threshold only indication in alert log
Posted on 2006-03-23
Oracle 10g on AIX with 6 started databases locked up. Users in session could continue processing but no new users could sign in. SQLPLUS froze after password was entered even when trying to sign in as sysdba.
Alert log showed at time of lock up a message "heap (2949) exceeds notification threshold (2048)". Appeared to be trying to load an object into SGA with name like XDB.xx.
It is possible that an import was running updating tables that one or more users could have been accessing, but alert log indicates successful import.
In trying to resolve problem, tried to stop listener to prevent attempted sign ins,
but "lsnrctl stop" also hung. Seemed like databases were down, but in trying to
strart databases, got messages to stop them first, so, depending on where you looked and what you did you got conflicting information as to state of databases.
Finally did brute force dbshut and dbstart.
Why I am getting heap size exceeded messages, and what is causing them?
Could whatever caused the messages cause an apparent lock up?