Solaris

Hi all,

Q1) Due to application requirement, i am required to have more swap space.
Currently my swap is on a partition with 32GB.
I have another partition with 100GB, but it already has a UFS filesystem on it.

Can i just swap -d /dev/dsk/current32gb and swap -a /dev/dsk/ufs100gb ?
Will there be a problem if the 100gb already has a filesystem on it ?

p.s. Due to personal preference, I do not want to create a swapfile and add on to the existing swap space.

Q2) The swap size recommendation for the application is to have 75% * RAM (128gb) = 96GB swap.

Do i really have to create the swap in multiples of 32GB (e.g.. 32, 64, 96) ?

Is there any performance impact if i just use a 100GB swap instead of a 96GB one ?

Regards,
Noob

Q1) Problem: the swap would trash the filesystem!

BTW my personal thumb formula is: swap-GB = 4 x SQRT(RAM-GB)

Q2) No, can be any size.

Q3) The larger the used swap, the lower is the paging to/from the swap. The unused swap does not matter for performance (but is wasted disk space).

Hi MadeInGermany

Thanks for your reply.
Q1) Do you mean that there will be a problem if i use a partition with an existing UFS filesystem as the swap device ? or do actually you mean "No problem" ?

Q2) That would be 4 x SQRT(128GB) = 4 x 11.31 = around 48GB ?

Q3) Can you elaborate abit further here ? - The "used" swap here means "allocated" swap or really "utilized" swap. I thought when memory is insufficient, paging to/from swap will occur. So how does having a big swap reduce paging ?

Last but not least, is it okay for me to use my 100GB partition with UFS filesystem as my new swap device ? - as I do not want to re-partition the harddisk and affect the rest of my partitions

p.s. initially i allocate 32GB of swap for my 128GB ram, but oracle's grid installation for database suggest 0.75(RAM) for swap, and hence i am trying to use an existing partition (100GB) as the actual swap device

Hope to hear from you soon.

If you use the same disk partition for swap and for a UFS filesystem, the contents of the UFS filesystem will be destroyed if any data is swapped to that swap space.

If you were willing to create a swapfile in the UFS filesystem and add that to your swap space, that would be different. But since you have said that is not an option, you MUST NOT allocate the same partition to be used as swap space and as an active filesystem.

Hi Don,

Thanks for your reply. The current ufs partition is use as a standby and i am willing to lose its current content.

Does that means I can use it to as a swap device despite having created an ufs filesystem on it; (as it will be overwritten during page swap) right ?

Ensure that the 100GB is not mounted. Then it can be trashed, you can add it to swap.
(If it's a partition on a physical disk on Sparc, it might give an error - in this case you must change the "wm" flag to "wu" in "format"s partition menu, in order to declare it unmountable. It's a good habit anyway.)
With used swap I mean the utilized swap, (blocks - free) in swap -l

Given the fact using a swapfile would be the most sensible solution, can you elaborate on this personal preference against it?

How much free space is available on the 100GB partition and on the root (/) one?

Will the application make actually use of the swap, i.e. is there a risk the application do more than reserving that extra memory and effectively write pages there?

If you are sure the answer is no for the last question, you might fake swap space by using a sparse file:

e.g.:

Code:

mkfile -n 64g /var/tmp/64g-sparse
swap -a `lofiadm -a /var/tmp/64g-sparse`