Re: Performance degradation

In article <Lur02.308$Dm.8889739_at_news.ipass.net>,   "David Sisk" <davesisk_at_ipass.net> wrote:

> Just a thought, here. High variability in response times usually indicates
> a general level bottleneck. Things to check would be: SGA hit ratios, O/S
> paging, disk workload (ie. reads or writes/sec), CPU utilization, rollback
> contention, redo buffer contention, redo log disk hot-spot,
> checkpoints/logswitch, freelist contention, and all the myriad other latch
> contentions.

> abeal_at_us.ibm.com wrote in message <71qmcc$66r$1_at_nnrp1.dejanews.com>...
> >We have an Oracle database running on Solaris 2.6 using Oracle 7.3.3 with
> >typically 1100 concurrent connections many of which are idle most of the
> >time. However, recently we started experiencing severe database
> degradation
> >where the database slows to a crawl. One of the DBA was able to observe a
> >high number of enqueues, library cache latches, and high number of shared
> >pool latches. When the degradation happens, it is usually so severe that
> >they have to shutdown the database and restart it.

Last night it happened again and he is what I observed. vmstat shows the CPU to be practically idle most of time (16 cpu sun box), no paging, little IO, and low load average. The oracle processes were not gaining much CPU time - I would expect the Oracle database processes to be very busy and the CPU to be very busy. There were about 600 connections and it took forever to login to the database - so this makes it hard to diagnose the problem. It was almost as though Oracle was spending lots of time internally waiting on something.

One question, would looking at syslog output or truss output on the database processes be of any help? -- Alan Beal

-----------== Posted via Deja News, The Discussion Network ==---------- http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own Received on Fri Nov 06 1998 - 07:53:25 CST