maxswapchunks = swchunk in 11.22 v1.6 for Itanium?

[Tim Lank_2]

I had accidentally posted this in the Ignite forum but it belongs here....

When I try and build the 11.22 v1.6 kernel on this rx5670 box, it gives the following message:

root@tybu8520 > pwd
/stand/build
root@tybu8520 > /usr/lbin/sysadm/system_prep -s system
root@tybu8520 > mk_kernel -s system
WARNING: No such device, tunable, or cdio maxswapchunks.
Ignoring the following line.
maxswapchunks 16384
Compiling conf.c...
Loading the kernel...
Generating kernel symbol table...
root@tybu8520 >


Does anybody know if the maxswapchunks kernel parm is not available in this environment? A kmtune before and after doesn't list this parm.

Thanks,
Tim

[James Murtagh · Accepted solution]

Hi Tim,

A quick look at the release notes does indeed mention this parameter has been obsoleted:

http://www.docs.hp.com/hpux/onlinedocs/5187-0701/5187-0701.html

Its under section:

HPUX Features
HPUX Kernel Configuration

Obviously it has had no impact on the kernel build, but a file in /usr/conf/master.d, no doubt core-hpux still has a reference to this. I wouldn't remove this just to be safe.

Regards,

James.

[James Murtagh]

Hi Tim,

Sorry, not Kernel Configuration, should be Tunable Kernel Parameters.

Regards,

James.

[Michael Tully]

try this document:

http://docs.hp.com/hpux/onlinedocs/TKP-90203/TKP-90203.html

As stated in your post that I had placed a reply, I believe the parameter your after instead of 'maxswapchunks' is 'swchuck'

[Tim Lank_2]

This is for an Oracle Database server. We've always set this in the past for PA-RISC boxes. As James Murdoch states, it appears from the doc that its obsoleted in 11.22 v1.6.

Do you still suggest setting the swchunk parm?

[James Murtagh]

Hi Tim,

I hope you have a film/TV theme joke going on here with my name! First it was Lethal Weapon with "Murtaugh" then "Murdoch" from the A-Team!! :-)

Anyway, I haven't looked in-depth at this, but it now looks like hpux uses a two stage table to allocate swap chunks. The second appears to be dynamically allocated making maxswapchunks redundant. I wouldn't change swchunks, see the man page for details, availave here:

http://www.docs.hp.com/cgi-bin/onlinedocs.py?mpn=B3921-90010&service=hpux&path=00/22/2211&title=HP-UX%20Reference

Under the memory paging section.

BTW, I just monitor the hpux section where I can see all posts put under it, Ignite included, so as Michael mentioned your post came up twice here. Its good to see you updating the thread with the info though!

Regards,

James.

[Tim Lank_2]

James,

Wow, I'm really sorry about butchering your last name - definitely must have had the movies/tv on the brain. I'll have the DBA double-check with Oracle on the swchunk parm before changing it...unless someone here recommends to change it.

Thanks!
Tim

[James Murtagh]

Hi Tim,

I actually thought you had planned that and it was really clever! Anyway, don't worry, happens all the time.... ;-D

Anyway, I have a few other threads I need to research when I'm back at work tomorrow....if I find any additional information on this I will be sure to add it to this post.

Take care.

James.

[Bill Hassell]

You can ignore any recommendation from Oracle or the DBA about swchunk. SAM has fairly strong warnings about changing it's value. This parameter sets the size of swap I/O operations and had nothing to do with Oracle. maxswapchunks simply enables or limits the maximum amount of swap space. In Itanium, the code has been heavily changed so there is no need to specify the maximum number of swap chunks that be created.

Swapping (actually, paging) is a very bad thing for database engines like Oracle. The very large swap recommendations apply only to installation--unless you have no RAM (less than 2Gb) in which case performance is of no concern to your users (it will be pitiful). If you have lots of RAM, you don't need a lot of swap. Swap space is not some magic enabler to make Oracle run...it is simply a storage area for pages of programs that are inactive. If the programs are all waiting on keyboard input, then this paging is fine. But if the paging rates are in the hundreds (vmstat, look at the po column) then your system will run very slowly. The ony fix is to add a lot more RAM.

[Tim Lank_2]

Based on Bill's post I believe this question has been fully answered. Thanks to everyone for their input.

--Tim