Paging Space issue!!

[Chris Fung]

Hi,

My machine has 2GB of memory and 4GB paging space. Unfortunately, the 4GB paging space is composed of two paging devices and sit on the same mirrored hard disks (we have 2 disks and mirrored). The default paging space 2GB and the other 2GB is user defined. Right now, I have a problem to add more hard disk to spread the 2 paging devices. As such I am planning to deleted the user defined paging space and then enlarge the default paging device to 4GB. One thing I am concerning is that I may not have enough continuous disk blocks to accomodate the extended paging space. Appreciated it if anyone can advice me alternatives to this issue.

Cheers,

Chris,

[S.K. Chan]

Assuming both your device swap have the same priority (hence interleaving is ON), it's not going to make any difference if you delete one of your swap device and extending the other to 4GB because it is still going to reside on the same disk. Based on your question you said to wanted to spread them on different disk, right ? My apologies, maybe I do not quite understand your question. When you say "user defined" paging space do you mean "secondary device swap" or "secondary filesystem swap" ? Can you also post the output of ..
# swap -tam
# lvlnboot -v
# vgdisplay -v
That will give a better understanding of your system configuration. Thanks ..

[James R. Ferguson · Accepted solution]

Hi Chris:

If I understand correctly, you have no spare physical disk, a mirrored boot volume (vg00), a 2GB primary swap (lvol2) and a 2GB secondary swap (lvolN), and want to have the best configuration for performance.

It is going to be virtually impossible to enlarge the primary swap (but see below). As you noted, the principal issue is finding physical disk extents that are contiguously available atthe "end" of the existing lvol2. This is not going to be the case since primary swap (lvol2) is laid down before the root ("/") directory (lvol3) during installation.

If you want to get around this, you best choice would be to make an Ignite recovery tape ('make_tape_recovery') of your system, and reinstall the server using the Ignite advanced installation menus. By doing this, you would be able to resize your primary swap (and eliminate the secondary one) during the installation process.

Remember, that if you choose to follow the Ignite method, you will need to manually remirror your logical volumes afterwards. Otherwise:

# make_tape_recovery -x inc_entire=vg00 -I -v -a /dev/rmt/0mn

The '-I' option of 'make_tape_recovery' is the *same* as the '-i' option of make_recovery. '-I' means cause the Ignite process to be interactive when booting from tape.

Another choice, however, is to set the swap priority of your primary and secondary swap to *different* values. Make one (the primary, for example) a higher priority than then other. Since I/O will *not* be interleaved in this arrangement, but rather one area will be preferentially used over the other until the other is needed, you will minimize disk mechanics (disk head movement). Remember, the lower the value for swap priority, the higher the bias.

Regards!

...JRF...

[Stefan Farrelly]

if you only have 2GB of RAM you only need 2GB of swap. I would delete the 2nd 2GB swap area leaving only the 2GB primary.

If you tried to extend the primary 2GB of swap it would fail. Youre right - you do NOT have enough contiguous space to extend primary swap - this would be extremely difficult to do, and as you dont need to extend it becuase you already have 2GB of swap (primary) and 2GB of ram, dont wast your time trying.

[James R. Ferguson]

Hi (again) Chris:

I also meant suggest, that if you don't already have the kernel parameter 'swapmem_on' that you should. If you don't have this turned on (the default) then you may find that doing so obviates the need for your 2GB of secondary swap in the first place!

Have a look at 'swapinfo -tam' for your current swap utilization. The presence of the line with "memory" indicates that psuedo swap ('swapmem_on') is enabled.

Pseudo swap doesn't really exist. Enabling it simply "fakes" the OS into thinking it does, allowing additional processes to start when all of the physical swap has been reserved. In fact, the kernel will consider you to have 75% of your physical memory added to the available swap pool. In your case that's almost the amount of your secondary swap.

Regards!

...JRF...