Running @ wrong NICE level

[Mike Blatsos]

I am running a scheduling package called $universe on 6 of my HP UNIX servers (11i OS) and on 4 of the servers the processes started by the scheduler are at NICE 39 on the other 2 they are at 20?? The scheduler processes are running at NICE 20 on all of the servers and the scheduler is started by root. I believe that the scheduler is run in the K shell and the processes it is running are in the C shell. The scheduler weekly stops and starts our dB (Oracle) and they are brought up at NICE 39. I have been told that Oracle does no like RENICE so I have to stop and restart Oracle from the command line. Has anyone seen this problem before?

[Steven E. Protter]

On an idle system dedicated to Oracle, oracle will probably struggle for and get what resources it needs with a NICE of 39.

But on a realistic system with some queries going on Oracle is not going to be happy.

This however is probably a problem with the scheduler, not the OS. I would check with the vendor for a bug fix or report the problem there.

There is a wonderful process scheduler that we use to stop oracle, back it up and turn it back on after cold backup.

Its called cron. We run the scripts with the su -c oracle "commandline" methodology. It works great. Certainly the price is right.

:-)

SEP

[Tony Willis_1]

It is my belief that all a unix scheduler does is create situations like this.
I agree that is VERY unlikely to be an OS problem.
The quickest way to tell is to use cron or even at from the command line.
This is pure OS and will elim the schduler.

Since we write scripts daily to cover repetative work, cron is a perferred tool here.

Tony

[Jeff Schussele]

Hi,

The OS by itself does not do any renicing. All it can do is adjust the run queue priority.
The only ways a proc can get a nice of 39 is:
1) The shell that spawned it was @ 39 or at 35 & spawned it the background.
2) The PID owner reniced it.
3) root reniced it.

That's all. I'd bet heavily on #1.

My $0.02,
Jeff

[Sridhar Bhaskarla]

Hi Mike,

It's not a OS problem. Any process that is started in background will be defaulted to a high nice (low priority) value.

I do not believe that oracle does not like to be reniced.

-Sri

[Mike Blatsos]

I have done more research and found that the scheduler pieces running in the K shell are running at 39, these are the pieces, which spawn the other jobs (i.e. Oracle, Omniback). I have little experience with the K Shell and don't know if background jobs run at 39, I did check the setup of ksh (ksh -version) and found bgnice off.

Note: We were running this in cron and it ran fine.

[Jeff Schussele]

Hi (again) Mike,

By default ALL processes run @ a nice level of 20. Any process spawned in the background is assessed a nice penalty of 4 - so a normal process started in the background would have a value of 24. So if the shell running these procs has a nice of 39, then it had to be started @ 39 - EX:

nice -19 ksh

HTH,
Jeff