Mark Brinsmead wrote,on my timestamp of 29/07/2006 11:55 AM:
But here's the half where you're wrong. (Or at least, not completely "right".) Those hypothetical
"other" applications can just as easily be resposibly for the *increases* in the slabinfo stats as
they are they would be for "non-zero" values. After all, those values had to increase from zero
*some* *time*, didn't they. ;-)
sure, but the chance of any of them touching the count when you start an Oracle instance is remote. Remember that I said to watch the counter WHEN you start Oracle.
The easiest way for me is to use the native "watch" linux program on cat /proc/slabinfo: it's not that big anyway and kio shows up around line 22 or so on mine: watch -n 2 cat /proc/slabinfo as root does the trick nicely.
Then you can see the cb counter (second column IIRC) ticking over when Oracle starts. Or not of course, if no aio is present!
As reliable as it'll ever get, I reckon. Short of digging into kernel memory and pulling out the actual cbs and their pointer links to each process.
Why on earth hasn't Oracle provided an easy and reliable method of finding this out (other than through the indirect async io statistic) is baffling. I think the problem is in every *n*x port, btw. Not just Linux. In fact, Linux is probably the only one where it is relatively easy to (non)confirm aio via the slabinfo trick: I'm not aware of an equivalent in Unix, short of strace? And that is not even very reliable...
-- Cheers Nuno Souto in sunny Sydney, Australia dbvision@xxxxxxxxxxxx -- http://www.freelists.org/webpage/oracle-l