Question of lsf (load sharing facility)

Question of lsf (load sharing facility)

Post by l.. » Thu, 05 Aug 1999 04:00:00



Hi,

I have a conceptual question to ask about Platform Computing's lsf (load
shring facility.

I have a server which acts as batch server. I need to allow user to
telnet to this server also.

Since the telnet user can run background jobs directly on the server,
those jobs will compete with the batch queues for the cpus on the
server. It seems that it is unfair to the queues users as they are
waiting quietly in the queue for CPUs but others can use the CPUs
without queueing.

Since I'm new to lsf, could any one advise me how to solve this
situation.

Thanks a lot.

Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.

 
 
 

Question of lsf (load sharing facility)

Post by Peter Ben » Thu, 05 Aug 1999 04:00:00



>I have a conceptual question to ask about Platform Computing's lsf (load
>shring facility.

>I have a server which acts as batch server. I need to allow user to
>telnet to this server also.

>Since the telnet user can run background jobs directly on the server,
>those jobs will compete with the batch queues for the cpus on the
>server. It seems that it is unfair to the queues users as they are
>waiting quietly in the queue for CPUs but others can use the CPUs
>without queueing.

>Since I'm new to lsf, could any one advise me how to solve this
>situation.

You could set the users' shells in the password file to wrapper
programs which reduce the resource limits and priority of the process,
then exec the real shell. Be careful not to set the limits too low,
or you might end up with truncated files when you do rcp/ftp on a
large number of files.

Provided that LSF never looks at the shell listed in the password
file, it'll never be affected by the reduced resource limits.

Peter

 
 
 

Question of lsf (load sharing facility)

Post by l.. » Fri, 06 Aug 1999 04:00:00






> >I have a conceptual question to ask about Platform Computing's lsf
(load
> >shring facility.

> >I have a server which acts as batch server. I need to allow user to
> >telnet to this server also.

> >Since the telnet user can run background jobs directly on the server,
> >those jobs will compete with the batch queues for the cpus on the
> >server. It seems that it is unfair to the queues users as they are
> >waiting quietly in the queue for CPUs but others can use the CPUs
> >without queueing.

> >Since I'm new to lsf, could any one advise me how to solve this
> >situation.

> You could set the users' shells in the password file to wrapper
> programs which reduce the resource limits and priority of the process,
> then exec the real shell. Be careful not to set the limits too low,
> or you might end up with truncated files when you do rcp/ftp on a
> large number of files.

> Provided that LSF never looks at the shell listed in the password
> file, it'll never be affected by the reduced resource limits.

> Peter

Hi,

However, it is easy for the user to reset the cpu limit since the hard
limit is of cputime unlimited in Solaris and cannot be changed via limit
-h (csh)

Finally, I'm planning to write a background job checking script which
will check and kill any background jobs running at night which are not
submited via lsf

Rgds.

Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.

 
 
 

Question of lsf (load sharing facility)

Post by Chris Thomps » Fri, 06 Aug 1999 04:00:00


[...]

Quote:

>However, it is easy for the user to reset the cpu limit since the hard
>limit is of cputime unlimited in Solaris and cannot be changed via limit
>-h (csh)

Where did you get that strange idea from?

  $ csh
  hostname.example% limit cputime
  cputime         unlimited
  hostname.example% limit -h cputime
  cputime         unlimited
  hostname.example% limit cputime 5m
  hostname.example% limit cputime
  cputime         5:00
  hostname.example% limit -h cputime
  cputime         unlimited
  hostname.example% limit -h cputime 6m
  hostname.example% limit cputime
  cputime         5:00
  hostname.example% limit -h cputime
  cputime         6:00

csh only stops you setting the hard limit lower than the soft limit. It also
lacks a syntax for setting hard and soft limits at the same time.

There are lots of good reasons for not using csh, but this isn't one of of them!

Of course, cputime limits are circumventable by individual processes anyway
in Solaris 2, just by ignoring SIGXCPU.

Chris Thompson
Email: cet1 [at] cam.ac.uk

 
 
 

Question of lsf (load sharing facility)

Post by l.. » Sat, 07 Aug 1999 04:00:00






> [...]

> >However, it is easy for the user to reset the cpu limit since the
hard
> >limit is of cputime unlimited in Solaris and cannot be changed via
limit
> >-h (csh)

> Where did you get that strange idea from?

>   $ csh
>   hostname.example% limit cputime
>   cputime         unlimited
>   hostname.example% limit -h cputime
>   cputime         unlimited
>   hostname.example% limit cputime 5m
>   hostname.example% limit cputime
>   cputime         5:00
>   hostname.example% limit -h cputime
>   cputime         unlimited
>   hostname.example% limit -h cputime 6m
>   hostname.example% limit cputime
>   cputime         5:00
>   hostname.example% limit -h cputime
>   cputime         6:00

> csh only stops you setting the hard limit lower than the soft limit.
It also
> lacks a syntax for setting hard and soft limits at the same time.

> There are lots of good reasons for not using csh, but this isn't one
of of them!

> Of course, cputime limits are circumventable by individual processes
anyway
> in Solaris 2, just by ignoring SIGXCPU.

> Chris Thompson
> Email: cet1 [at] cam.ac.uk

Hi,

Thanks for your reply. I got this idea from Essential Sys Admin
(O'Reilly) but I have learned that it works for Solaris from you

Thanks

Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.

 
 
 

1. problems configuring shared memory facility and semaphores facility

I suppose I needs to add this to the FAQ.

Anyway, semaphores aren't loaded until needed.  Ipcs is broken in that
it gives info on whether a system is loaded or not, not whether it
is available.  If you use them, you'll find the semaphores are there.

Casper
--
Casper Dik - Network Security Engineer - Sun Microsystems
This article is posted from my guest account at the University

Opinions expressed here are mine (but you're welcome to share them with me)

2. general mail lock ?

3. Load Balancing Facility

4. AMD PCI SCSI controller in HP Vectra 5/90 XU

5. segmentation fault ?

6. wintern: toward a shared desktop web cache facility

7. NIS+ user management [Was: Re: root changing a user's password (NIS)]

8. error in loading shared libraries: libXmu.so.6: cannot open shared object

9. Shared library loading shared library.

10. Shared memory facility & Solaris 2.3

11. shared memory facility

12. Shared variable between main() and loaded shared library - dlopen()