LINUX Kernel Parameter

[Jojo Castro]

hi All,

Need your inputs on how can i monitor kernel parameter usage for LINUX (version 4 and up)?

Thanks!

[Zinky]

It all depends on WHAT kernel parameters you are concerned with sir.

I use HP GlancePlus as it is our standard accross UNIX dialects and Linux distributions.

[brian_31]

Alzhy

We use RHEL5, is there a package for Glance for RHEL5?

Thanks

Brian.

[Zinky]

Nope.
It's commercialon-free from HP.

You can however install a TRIAL version which should be good for 90 days(?)

[Jojo Castro]

Its Linux not HP-UX. For example i want to know the current utilization of kernel param shmmax, how can i check it?

Thanks!

[Marcin Bartoszek]

Hi,

Usually you can find actual kernel params in /proc/sys/kernel.

Particularly, you can check SHMMAX value by:
# cat /proc/sys/kernel/shmmax

Regards,
Marcin

[Jojo Castro]

Hi Marcin, I know this.
My question is super clear, how can i monitor the usage?

[Marcin Bartoszek]

Exactly what action you would like to monitor?
Operations of reading or changing kernel parameters?
Utilization of kernel resources?

Serveral reports and statistics you can get using tools like sar or vmstat.
You can also use collectl from sourceforge - I think this should be available as RH rpm package.

Are you looking for a commercial tool, or rather wondering how to implement it yourself?

M.

[dirk dierickx]

shmmax is a setting where you specify the maximum size of a memory segment, it is not something you monitor like cpu or overal memory usage.

with the command 'ipcs -m' you can view all the segments in use.

[Jojo Castro]

Hi Marcin,

Exactly what action you would like to monitor?
==> Utilization of kernel resources

You can also use collectl from sourceforge
==> what does this utility do?

Are you looking for a commercial tool, or rather wondering how to implement it yourself?
==> not exactly a commercial tool since our box is already in production.

Hi Dirk,

Ok its "with the command 'ipcs -m' you can view all the segments in use."

Ok just to add and be more specific with my question,
For LINUX boxes with DB and Oracle client running on it, what are the kernel parameters I should closely monitor? and How should i monitor them?

Answer like that of Dirk:
shmmax => ipcs -m

[Marcin Bartoszek]

You can find collectl and doc here:
http://collectl.sourceforge.net/

M.

[Zinky]

Jojo Sir,

Glance is universally available for most UNIX and UNIX like platforms sir -- Linux included of course.

Using it is a good skill set to have.

And yes you can monitor most kernel params with it.

But to answer thine question on what kernel parms to monitor (or set properly) for Oracle -- well aside from teh recommended settings for shmmax and various SHM/SEG tunables - I pay particular attention to:

nproc (monitor vis plain ole ps)
nofile (monitor via lsof)

ipcs -l (to monitor current settingd for SHM and SEM)

kernel.sem SEMMSL SEMMNS SEMOPM SEMMNI

SEMMNS - value should be greater than the total of "ipcs -s" nsems column. Generally it should be 1.5x than the sum of all Oracle DB's PROCESSES params. You will know you are getting close to kernel SEMMNS limits by simply doing a count of all Oracle processes (background and foreground -- simple ps)

HTH .

[Jojo Castro]

thanks

[dirk dierickx]

i just want to add that you will find these parameters in the guides of the oracle version you are using.