Who takes 99% of CPU
Posted in 2000
Topics: General Discussion
Hi guys,
When i do a ps aux , i see that the online process is running at 99% on the
CPU.
I would like to know if there's a way to know who is taking all the CPU
in that instance. onstat "something".
Because when i "onstat -u" nothing moves (no read no writes),
maybe it's a calculation process.
Thanks
pfiset@videotron.net
Pascal Fiset wrote:
> Hi guys,
>
> When i do a ps aux , i see that the online process is running at 99% on the
> CPU.
>
> I would like to know if there's a way to know who is taking all the CPU
> in that instance. onstat "something".
>
> Because when i "onstat -u" nothing moves (no read no writes),
> maybe it's a calculation process.
Do you have a NETTYPE parameter in your ONCONFIG file that uses a CPU VP
as a listener for a TCP/IP connection? This will cause the main CPU VP to poll
the
network connection continuously when it is not busy to see if there are any
requests
pending that it can work on. ALWAYS use NET VPs for TCP/IP listeners and CPU
VPs for shared memory listeners to prevent this. If this is the case the VP
will be
using mostly system time and the number of system calls per second will be
immense,
if you have a monitor that can show them.
Art S. Kagel
Pascal Fiset wrote:
> Hi guys,
>
> When i do a ps aux , i see that the online process is running at 99% on the
> CPU.
>
> I would like to know if there's a way to know who is taking all the CPU
> in that instance. onstat "something".
>
> Because when i "onstat -u" nothing moves (no read no writes),
> maybe it's a calculation process.
>
> Thanks
> pfiset@videotron.net
onstat -u reports nreads and nwrites. These correspond to the Pagreads andPagwrites (sysmaster:syssesprof) carried out on behalf of that session. A
session could be using bufreads to do its stuff, in which case "onstat -u"
would seem to indicate no apparent activity, even though cpu-intensive stuff
is going on using data available within BUFFERS. This is typical of badly
written SQL (correlated subqueries without supporting indexes, etc).
Examine sysmaster:syssesprof for rapidly increasing bufreads or isreads or
lockreqs (after inititializing stats, if required [onstat -z]) to determine if
this is the case, and, if so, which session(s) is your problem . Then, there
is an interesting command, "onmode -z ", that could be wielded...
Rudy