Re: oracle cause "load" on OS level very high.

Hi,

Thank you.
 Wow, 40 really means 40 processes in the RUNNABLE state.. I think I finally understand what you trying to tell me. I got it now :)

Dominica

"koert54" <nospam_at_spam.com> wrote in message news:<AOT49.56752$8o4.7773_at_afrodite.telenet-ops.be>...
> > 40 DOES not mean 40 processes.
> > Itis the "load" on Unix system.
> > Actually at that moment, it say it is running 970 processes in the
> > system.
> >
>
> hehe - a load of 40 *DOES* actually mean 40 processes in *RUNNABLE* state
> :-)
> and they are all queued up to run on your tiny CPU ... hence the name
> runqueue
>
> If you don't know what the process states runnable, sleep, zombie, pre-emted
> etc mean
> you can start by reading up on unix...
> may I suggest the following books :
> "The Design of the Unix operating system by Maurice J Bach"
> http://www.amazon.co.uk/exec/obidos/ASIN/0132017997/ref=sr_aps_books_1_1/202
> -3322413-4230252
> Well known excellent book - must read !
> "Optimizing Unix for performance by amir h majidimehr"
> http://www.amazon.co.uk/exec/obidos/ASIN/0131115510/qid%3D1028917178/202-332
> 2413-4230252
> Unknown book but it kicks ass - Unix DBAs -> buy it !
>
> But of course - you could always start by typing "man uptime" or "man top"
> at your prompt and it'll explain *load* for you :
> "The uptime command prints the current time, the length of time the system
> has
> been up, the number of users online, and the load average. The load average
> is
> the number of runnable processes over the preceding 5-, 10-, 15-minute
> intervals. The output of the uptime command is, essentially, the heading
> line
> provided by the w command."
>
>
>
>
> "Dominica Leung" <dominica_l_at_yahoo.com> wrote in message
> news:5f8af180.0208081730.1ccc1c85_at_posting.google.com...
> > Hi,
> >
> > 40 DOES not mean 40 processes.
> > Itis the "load" on Unix system.
> > Actually at that moment, it say it is running 970 processes in the
> > system.
> >
> > You know what, I am NOT USING MTS,
> > but I should,
> > let me ask you this, if
> > I have 970 processes, how many dispatches should I use??
> >
> > You are right, it could has so many FACT to have load that high.
> > But you know what, I think this company is really MAX out this
> > TINY POOR little machine.
> > They do 582 select per second. It is a lot.
> > Thanks for your help tough.
> >
> > my next project is to setup CONNECTION manager, I don't know how yet.
> >
> > Dominica
> >
> >
> > ==============
> >
> > "koert54" <nospam_at_spam.com> wrote in message
> news:<4Qd49.53393$8o4.7093_at_afrodite.telenet-ops.be>...
> > > if I'm not mistaken the 'load' is the number of runnable processes ...
> so if
> > > it peeks to 40 it means
> > > 40 processes are runnable at that moment ... this could mean that :
> > > - someone is using parallel query (pq slaves) ?
> > > - someone wrote some stupid program that connects multiple times
> > > concurrently is a short amount of time ?
> > > - dbwriter is using dbwriter slaves ? multiple arch slaves ?
> > > - a process is spawning a hugh amount of threads ?
> > > - combination of above ...
> > > - other reason :-)
> > >
> > > My experience is on boxes that show very high runqueues and peaks of
> high
> > > runqueues, PQ is being abused - parallel processing
> > > is not always faster ... one time a saw a box with a runqueue (load) of
> 80
> > > to 100 ... reason was a high parallel degree clause on most tables -
> > > lowering PQ -> low runqueue -> lower overhead, context switches ->
> increase
> > > performance ...
> > > My experience is, depending on number of CPUs, a runqueue/load of 2 to 3
> > > times number of CPUs is a performance limiet...
> > >
> > > "Dominica Leung" <dominica_l_at_yahoo.com> wrote in message
> > > news:5f8af180.0208061359.6cac32d6_at_posting.google.com...
> > > > Hi All,
> > > >
> > > >
> > > > I am wondering whether anyone seen this kind of situation
> > > > before in Performance tuning aspect.
> > > >
> > > > I am (this company AA) is running oracle 8.0.6.2
> > > > 32 bits on
> > > > solaris 2.6 OS.
> > > >
> > > > There is very strange thing happen.
> > > > When I do a "top",
> > > > everyday at 2:30 pm and 2:45PM
> > > > (there are some cronjob from Engr dept doing select and
> > > > update)
> > > > the "load" from the "top" command, will be 40.00.
> > > > This is very high, normal machine is 4 or 5.
> > > >
> > > > =========================================
> > > >
> > > > last pid: 8806; load averages: 40.00, 8.57, 9.67
> > > > 14:50:39
> > > > 928 processes: 923 sleeping, 2 running, 3 on cpu
> > > > CPU states: 32.9% idle, 54.0% user, 13.0% kernel, 0.0% iowait, 0.0%
> > > > swap
> > > > Memory: 4096M real, 133M free, 282M swap in use, 1254M swap free
> > > >
> > > > =================================================
> > > > It only spike up for 3 or 4 minutes.
> > > >
> > > > When I do top, I only see the
> > > > PID is 25923, which is
> > > > the oracle 25923 1 0 Jun 29 ? 1031:47 ora_dbw0_USRDEV
> > > >
> > > > My question is ?
> > > > 1) is there anything I should look??
> > > > I have looked at V$lock, v$transaction at that time,
> > > > there is NO PROCESS waiting for transaction, only 14 locks in the
> > > > whole system.
> > > >
> > > > Curious how to improve it, look like database writer is very busy.
> > > >
> > > >
> > > > I know people might told me "this machine is MAX out" in the
> > > > capacity. But I hope there is something I could do about it.
> > > >
> > > >
> > > > Thanks,
> > > >
> > > >
> > > > Dominica
Received on Mon Aug 12 2002 - 15:44:01 CDT