Very Computer

the default value is 15000 I want to go up to 30000 because my server has
become so slowly.
I get the error : 00:2001/05/10 15:07:26.40 kernel  os_create_region: can't
allocate 58044416 bytes

It is SQL Server/11.0.3.3/P/Linux OS

here is a print of my 'TOP' :

  6:05pm  up  1:01, 11 users,  load average: 0.11, 0.16, 0.14
99 processes: 95 sleeping, 4 running, 0 zombie, 0 stopped
CPU states:  5.2% user,  1.1% system,  0.0% nice, 93.9% idle
Mem:  257088K av, 215200K used,  41888K free, 108168K shrd,   6148K buff
Swap: 130748K av,      0K used, 130748K free                 77220K cached

any ideas ???

Thanks a lot,
Xavier.

Please see the Sybase on linux FAQ at
http://www.mbay.net/~mpeppler/linux.html

Michael
--


International Sybase User Group - http://www.isug.com

Doesn't change anything ! He even doesn't want 20000 !
I have kernel 2.0.36

I changed to :

#define SHMMAX 0x4000000 /* max shared seg size (bytes) */

but no way ...

any hints ?
by the way, any opinion about segmenting the default data cache which has
only a pool size of 2K in several with 4K, 8K or even 16K ?

Thanks,
Xavier

Quote:> by the way, any opinion about segmenting the default data cache which
> has only a pool size of 2K in several with 4K, 8K or even 16K ?

No doubt others here will have some comments, though.

Michael

    Michael> Houppertz"

    >> Hi,
    >>
    >> Doesn't change anything ! He even doesn't want 20000 ! I have kernel
    >> 2.0.36
    >>
    >> I changed to :
    >>
    >> #define SHMMAX 0x4000000 /* max shared seg size (bytes) */

Did you actually compile and install the new kernel?  This is not like some
Solaris where the kernel build is dynamic.

    Michael> Well - it's been a long time since I did it for a 2.0.36
    Michael> kernel, but the process outlined in the FAQ worked for me.

And me!

    >> by the way, any opinion about segmenting the default data cache
    >> which has only a pool size of 2K in several with 4K, 8K or even 16K
    >> ?

Well, the best place to go is the performance and tuning guide.  I don't
think that the basic principles changed between 11.0.3 and 11.5.1.  

A couple of quick pointers.  Don't have more than two pools (1 x 2k + 1
other).  The optimiser attempts large IO with the largest available pool
and then drops back to the 2k pool if that is unavailable.  Sizing of the
pool is problematic with that release since it is hard to see how well the
cache is being utilised without sp_sysmon.

dowen
--

Maintainer of the Sybase FAQ: http://www.isug.com/Sybase_FAQ

to set a new shared memory max setting runtime issue to command
sysctl -w "kernel.shmmax=<any number>"
and add the following line to /etc/sysctl.cfg to automatically enable
this setting upon next reboot
kernel.shmmax=<any number>

bjornw>
--
------------------------------------------------------------------------

Michael
--


International Sybase User Group - http://www.isug.com

Quote:> >> Did you actually compile and install the new kernel?  This is not like
> >> some Solaris where the kernel build is dynamic.

> > You don't have to hardcode this settings in the kernel anymore.

> > to set a new shared memory max setting runtime issue to command sysctl
> > -w "kernel.shmmax=<any number>" and add the following line to
> > /etc/sysctl.cfg to automatically enable this setting upon next reboot
> > kernel.shmmax=<any number>

> Err... the original poster talks about 2.0.36 - he definitely needs to
> rebuild his kernel...

Thanks,

Xavier

I myself is using the latest 2.4.xx kernels, because the SMP support is
far better than in 2.2 kernels. Note that performing backups of large
databases on the 2.4.[0-3] versions will probably end up eating up all
your CPU for an hour or so due to some buggy VM implementation. (seems
to be solved in 2.4.5 kernel) - (I can post more information about this
issue if other people are experiencing the same problem)

Both kernels have pro and cons, but they both share the abillity to
set shared memory as explained in my previous posting. So urge you
to upgrade the kernel - and remember to upgrade all other packages,
especially the modutils package to cope with the latest kernel)

bjornw>
--
------------------------------------------------------------------------

    >> >> Did you actually compile and install the new kernel?  This is not
    >> like >> some Solaris where the kernel build is dynamic.
    >> >
    >> > You don't have to hardcode this settings in the kernel anymore.
    >> >
    >> > to set a new shared memory max setting runtime issue to command
    >> sysctl > -w "kernel.shmmax=<any number>" and add the following line
    >> to > /etc/sysctl.cfg to automatically enable this setting upon next
    >> reboot > kernel.shmmax=<any number>
    >>
    >> Err... the original poster talks about 2.0.36 - he definitely needs
    >> to rebuild his kernel...

    Xavier> How do i recompile or rebuild the kernel ???  Seems dangerous ?
    Xavier> please give me the complete action to take cose I'm not a
    Xavier> * user of linux and I don't want to blow it out ...

Probably the best thing to do is go and read the Linux Admin Guide
(available from http://www.veryComputer.com/) and there may even be some more
documentation there as well.  If you grab a new kernel from
http://www.veryComputer.com/ it comes with detailed instructions on what to do.
You do not threaten the life of the server while building the kernel, it is
only installing and rebooting that is dangerous.  I would certainly
recommend practising this on a different box if at all possible, especially
if the system that you are using is in any way important to you.

One important tip:  Always leave a kernel in place that you know is safe,
with the corresponding lilo entries, so that you can recover.  The docs
will tell you *how* to do these things, it can seem like an easy timesaver
to skip this.

dowen
--

Maintainer of the Sybase FAQ: http://www.veryComputer.com/

In Solaris it's a config file, maybe under linux you need to recompile kernel.

Pierre

Quote:> Hi,

> the default value is 15000 I want to go up to 30000 because my server has
> become so slowly.
> I get the error : 00:2001/05/10 15:07:26.40 kernel  os_create_region: can't
> allocate 58044416 bytes

> It is SQL Server/11.0.3.3/P/Linux OS

> here is a print of my 'TOP' :

>   6:05pm  up  1:01, 11 users,  load average: 0.11, 0.16, 0.14
> 99 processes: 95 sleeping, 4 running, 0 zombie, 0 stopped
> CPU states:  5.2% user,  1.1% system,  0.0% nice, 93.9% idle
> Mem:  257088K av, 215200K used,  41888K free, 108168K shrd,   6148K buff
> Swap: 130748K av,      0K used, 130748K free                 77220K cached

> any ideas ???

> Thanks a lot,
> Xavier.

I downloaded RedHat 7.1 : can I just tell him to upgrade my linux ? Would i
loose something ? I have some specific boards like a multiport pci card, and
my apc ups, ... will I have to reinstall the or will he keep them working ?

Thanks,
Xavier

best of luck - and remember backup is for newbies :-)

--
------------------------------------------------------------------------