MUTEX errors -- Significant? How to fix?
85558Jul 27 2010 — edited Jul 27 2010I've started receiving notices from a database monitoring tool with the following information:
Warning: Mutex contention on ora2-prod is at 564.00 lock attempts per second. This signifies a multi-processing overload condition.
But I can't seem to track down what might be causing this issue (and it's happening more and more) nor how to handle it. Also, I can't seem to find any information on what a good threshold for triggering this error might be. Is "564.00 lock attempts per second" high? What's a "good" value? How high can it go before I hit the alarm for general quarters?
If it's significant, this particular server (ORA2-PROD) is the secondary node of a two-node cluster. Both nodes are running on identical Sun Solaris boxes. Any and all help gratefully appreciated.
Here is more information:
The tool reporting the error is Foglight, a third-party tool we use to monitor various aspects of our production operation. I have NO further knowledge about the tool because I don't actually use the tool, I just get the emails.
Our O/S is SunOS 5.10 running on a SPARC-Enterprise-T1000 ("Solaris 10 8/07 s10s_u4wos_12b SPARC")
Our version of Oracle is Oracle Database 10g Enterprise Edition Release 10.2.0.4.0
I'm still trying to find out what the current settings are for the semaphores on this system but I didn't set up this instance and I'm not the actual administrator. I'm just the person everyone looks at whenever anything unusual happens with the database.
The point to my question though is, how many "lock attempts per second" is considered significant. Is it more than 500 (which seems to be the threshold that's being monitored)? More than 740 (which is the highest that's been reported near as I can tell)? Or is the number unimportant and the mere FACT that we're HAVING multiple lock attempts being made "per second" significant?
The actual administrators (who live far off-shore in a time zone that's 14+ hours off from mine) suggested increasing the number of semaphores on the server. I'm just curious what the "best practices" are for this situation.