Hi,

Oracle 8.0.5, MTS, HP-UX 10.20: My v$session_longops is
monitored via a web page, updating every 10 seconds since the database
slows down so much at 2 pm, that I restart it with shutdown abort,
just to get it up again. The v$session_longops is checked to see what
people are doing at the time, to find any offensive sql's. I have
three entries in there at the moment for sessions with a status KILLED
(I did that yesterday). How do I clear them ?

Oracle 8.0.5, MTS, HP-UX 10.20: I'm running MTS on oracle
8.0.5, 30 mts_servers, max 50. Only 80 users, 18 running at the
moment, 12 idle, but those running are at 40% capacity. Can it not
utilise the other shared servers to drop the load? What else can I
look at? I'm running stats pack at the moment, but need to find where
the problem is in the application. I can link the shared_servers to
the session and longops to see why the servers remain high for so
long, but they may not be performing longops. Should I link to the
v$sqlarea to see the sql's being executed ?. How else can I dig into
the database to find what is going wrong ?

This slowdown occurs pretty much at the same time each day, so
can anyone suggest what else/how I should be seeking the problem. I do
not have access to the application source code, and therefore cannot
check the functionality from that way. Could be that the system simply
cannot keep up with the user load, but the sar output shows 360
processes from 1024 at the time of the failure yesterday, and file-sz
well below the threshold, sar -b shows box 90% idle, therefore I do
not suspect the system.

Any assistance would be greatly appreciated.

Cheers

GJC

==================================================================
This e-mail and its contents are subject to the Telkom SA Limited
e-mail legal notice available at
http://www.telkom.co.za/TelkomEMailLegalNotice.PDF

Oracle LazyDBA home page