Contiguous space

[Waugh]

Hi,

Is there any reason for swap space we r using
Contiguous space. OR why we r using Contiguous space for HP-unix installation on boot disk.

Regards
Rkumar

[Waugh]

Hi,

what will happen if i m not using Contiguous space

Regards
Rkumar

[Analyst]

Hi Rkumar,


NO,Its not recommended.

Set the contiguous allocation policy. A contiguous logical volume has three characteristics:
1.Physical extents are allocated in ascending order,
2.No gap is allowed between physical extents within a mirror copy,
3.Physical extents of any mirror copy all reside on a single physicalvolume.
Use the strict (-s) and contiguous (-C) options together to form various combined allocation policies on a logical volume. For example, -s y -C y defines a logical volume such that each mirror copy is contiguous, yet mirror
copies of a logical extent cannot share the same physical volume.
contiguous can have one of the following values:

y Set a contiguous allocation policy.

n Do not set a contiguous allocation policy. This is the default.


Thz,
Analyst.

[Matti_Kurkela]

Usually the primary swap does double duty as a crash dump area.

If the system crashes (for example, by a serious kernel problem or because TOC button was pressed), then the system firmware takes control of the processors and tries to create a crash dump to a designated crash dump area.

The crash dump routine is very simple: the only information it can take is "where does the dump space begin, and how big is it?" The firmware does not understand LVM nor VxVM and blindly overwrites anything in the dump area. It will copy the entire system memory, or as much of it as possible.

When HP-UX is restarted after the crash, the system will detect that a dump exists and tries to compress & store it to /var/adm/crash. This happens very early in the boot sequence, just before swap devices are activated for paging.

Also, the paging routine can work more efficiently if the swap space is contiguous. The HP-UX kernel could handle non-contiguous paging areas. Essentially, it does that already: if you define two or more swap devices, or any swap files in addition to the primary swap, your swap is split into two or more pieces and thus is not entirely contiguous.

But that is an optimization issue: a contiguous (or even mostly contiguous) swap device is simpler to use and thus more efficient.

MK

[Emil Velez]

The swap on the boot disk must be in a contiguous lvol or vxvm volume. That is because the kernel does not know how to understand lvm or VXVM.

Other swap that is configured after boot can be standard logical volumes. It does not seem to be a issue whether secondary swap not on the boot disk is contiguous or not.

In fact it can be a striped LVOL if the luns are from a disk array.

one issue to be aware of is that swap space must be protected or mirrored. I say protected meaning if the swap lvol is on a RAID disk you do not need to mirror it since it will not fail. If a swap device fails the system will panic sooner or later unless it is mirrored using LVM or VXVM.

By the way you also should not really need much swap space since you should be able to run everything in memory. Memory is relatively cheap and people can buy as much memory as their apps need.