Re: Mutex contention on database server

Hi, Dominica,

Virtual Adrian is a Solaris performance monitoring tool written by Adrian Cockcroft. That mutex contention field comes from `mpstat` output smtx column, defined as "spins on mutexes (lock not acquired on first try)". As Jonathan pointed out, these spins roughly translate to latch acquisition inside Oracle. The only thing your UNIX system admin can do may be replace the CPUs with faster ones (instead of adding more CPUs). But the root cause of the problem, if it's a problem at all, is tuning Oracle. For instance, if you see frequent "latch free" wait in v$session_wait and p2 indicates library cache, then make sure to use bind variables, etc.

Mutex contention is not related to semaphore setting in /etc/system.

Yong Huang Received on Sat Apr 30 2005 - 23:33:51 CDT