Operating System - HP-UX
1846856 Members
3541 Online
110256 Solutions
New Discussion

Re: Monitoring kernel parameters

 
Joe Profaizer
Super Advisor

Monitoring kernel parameters

Is there a method of monitoring or checking the actual used values of all or most kernel parameters. I understand that "glance -t" gives a select few, but I was wanting to see if we're pushing the limit on any kernel parameter on our system.

Thanks,

..Joe
16 REPLIES 16
Pete Randall
Outstanding Contributor

Re: Monitoring kernel parameters

Joe,

Glance is the only tool I know of that gives you high water mark, current usage, etc.


Pete



Pete
Bryan D. Quinn
Respected Contributor

Re: Monitoring kernel parameters

kmtune | more

This will show your existing kernel parameters.

-Bryan
doug mielke
Respected Contributor

Re: Monitoring kernel parameters

I'd like to add to this question. In another unix (SVR4?) you could run the 'crash' command and get this, or almost anything else from the kernel. Is there an hp equivalent? M4 mayby?
doug mielke
Respected Contributor

Re: Monitoring kernel parameters

I'd like to add to this question. In another unix (SVR4?) you could run the 'crash' command and get this, or almost anything else from the kernel. Is there an hp equivalent? M4 maybe?
Joe Profaizer
Super Advisor

Re: Monitoring kernel parameters

I don't believe kmtune will show actual highwater and current parameter utilization versus what's available.

Or does it w/ the correct parameters?

..Joe
Pete Randall
Outstanding Contributor

Re: Monitoring kernel parameters

Joe,

No, kmtune will just show you the current values.


Pete



Pete
Steven E. Protter
Exalted Contributor

Re: Monitoring kernel parameters

In combination with the kmtune output you can look at performance data you collect with sar and get some idea how much of certain kernel parameters you are using.

Its a manual process, not nice and graphical like glance, but you get more usable, tweakable data by using sar scripts.

Attaching some examples.

SEP
Steven E Protter
Owner of ISN Corporation
http://isnamerica.com
http://hpuxconsulting.com
Sponsor: http://hpux.ws
Twitter: http://twitter.com/hpuxlinux
Founder http://newdatacloud.com
William Wong_2
Trusted Contributor

Re: Monitoring kernel parameters

Itanium HP-UX or PA-RISC HP-UX? In the case of Itanium HP-UX you can get that information with kcmond. As for PA-RISC there are some unsupported tools that you might be able to get if you call the response center and ask for them.
Bryan D. Quinn
Respected Contributor

Re: Monitoring kernel parameters

Hey Joe,

No kmtune will only show you your existing values. Strictly a display of the parameters.

-Bryan
Bill Hassell
Honored Contributor

Re: Monitoring kernel parameters

Very few kernel parameters are values that you need to monitor. The ones that need watching are those which define the size of a table such as nfile and nproc. Both of these (the most common ones that need adjusting) can be seen with:

sar -v 1

Other parameters such as IPC tables (shared memory, semaphores, message queues) will require tools like Glance to monitor. Fence parameters (to limit runaway programs) don't fill up, they just prevent processes from consuming too much of the system resources (files, memory, etc).


Bill Hassell, sysadmin
Joe Profaizer
Super Advisor

Re: Monitoring kernel parameters

Here's the problem. We're trying to start up an oracle database, and we're getting the error, ENOSPC on a semget command.

Any clues to what parameter this may refer to?
William Wong_2
Trusted Contributor

Re: Monitoring kernel parameters

Ok semget refers to the declaration/creation of a semaphore. My guess is that it's one of the following kernel parameters:

msgmap
msgmax
msgmnb
msgmni
msgseg
msgssz
msgtql
semmni
semmns
semmnu
semume
shmmax
shmmni
shmseg

These are different kernel tunables for IPC.
Joe Profaizer
Super Advisor

Re: Monitoring kernel parameters

I found the below excert while searching the ITRC with "evospc" as the search phrase.

Here's something similar from an Omniback question, but I bet it's the answer for you to.

The error occurs because of "enospc / semget error in Oracle", semget error unable to get first semaphore set and no space on device. When the system does not have enough contiguous semaphores in the sem array to satisfy a semget request, the call returns a ENOSPC error. This error may occur even though there may be enough free semaphores, but they are not contiguous. Run "ipcs -a -s" and then add up the number under column "NSEMS". This sum should be the no. of semphores used. If this is the cause of the problem, increase kernel parameter "semmns" which is the total number of semaphores in system. Check semmns and increase semmns.
KCS_1
Respected Contributor

Re: Monitoring kernel parameters

hi,Joe

There is a monitoring tool of the kernel parameter what are you looking for.

Have a look at this that how to install and manage for using many kinds of funtions on this tool.

http://software.hp.com/products/KCWEB/downloads.html


Easy going at all.
malay boy
Trusted Contributor

Re: Monitoring kernel parameters

Joe,
How about sar?.I know sar -v can see the below :

14:11:33 text-sz ov proc-sz ov inod-sz ov file-sz ov
14:11:34 N/A N/A 125/2420 0 1224/2888 0 830/4779 0

regards
mB
There are three person in my team-Me ,myself and I.
Wouter Jagers
Honored Contributor

Re: Monitoring kernel parameters

I often use ktstat on my servers. It can check file, inode, lock (often very useful!) and process tables in your kernel.

You can get it at the porting and archive centre: http://hpux.connect.org.uk/hppd/hpux/Sysadmin/ktstat-1.42/

Regards
Wouter
an engineer's aim in a discussion is not to persuade, but to clarify.