We have four Oracle 9i instances on a Compaq Alpha running Tru64 Unix 5.1. If you shut down the Oracle listener while the databases are up, each pmon will spike and take up all of the available cpu and stay there. We are able to reporoduce this problem on more than one machine. On a four way machine with four instances, each pmon will take 99% of a cpu. On a two way machine with 4 instances, each pmon takes 50% of a cpu. The workaround has been to either a)not stop the listener while databases are active, or b) once PMON starts spinning, shutdown each database and restart. Anyone else run into this problem?
I also faced the same problem. I'm using Oracle 8.1.5 on AIX OS 4.3.3 - RS6000 IBM machine.
This machine comes with 4 CPU.
It all happens when the Oracle Listener hung. When I re-started the CPU, I noticed the pmon process is taking 100% on one of the CPU. This is strictly not acceptable. As such, I scheduled for a re-start on the database. I'm not sure is this a known problem and I really wish someone would have come across and let us know what need to be done.
1. Does anyone know why listener hang could affect pmon?
2. Can pmon be re-started by any other way other than re-starting the database?