On Fri, 2005-11-04 at 08:35, Andrei Dorofeev wrote:
> Hi Mike,
>
> > processes are counted multiple times. In an extreme example of a busy
> > 15k domain with hundreds of gigabytes of physical RAM and less than 50
> > GB swap, "prstat -a" will report that many terabytes of memory are in
> > the resident set for the oracle user. This is impossible.
>
> About the only thing prstat could do today about this is to read
> /proc/pic/map files and go through all memory segments for each
> process looking for processes using the same pr_shmid (see proc(4)).
> That will slow down prstat and yet won't not be 100% accurate as
> there are other types of shared memory (shared libc text, for example).
However, there is one case in which prstat could be improved. While
the prstat -a output simply adds the memory of each process, if you
use prstat -aL it simply adds things up by LWP so the memory usage
gets multiplied by the number of LWPs in the process. eg:
NPROC USERNAME SIZE RSS MEMORY TIME
CPU
2 apache 708M 534M 5.0% 1:27:31 6.7%
30 oracle 11G 9915M 93% 0:11:31 6.0%
50 root 138M 84M 0.7% 8:18:21 0.7%
NLWP USERNAME SIZE RSS MEMORY TIME
CPU
123 apache 84G 63G 55% 0:50:06 7.1%
156 oracle 59G 52G 45% 0:11:16 5.7%
122 root 447M 323M 0.3% 8:18:16 0.7%
In this case the apache user is running a single, but rather large,
java process.
(Just looking at this, I'm not sure why the TIME is so different.)
--
-Peter Tribble
L.I.S., University of Hertfordshire - http://www.herts.ac.uk/
http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/
_______________________________________________
perf-discuss mailing list
[email protected]
