Re: why the difference in swap value for vmstat output
From: Bill (bill12317_at_lycos.com)
Date: 11/14/03
- Next message: Dave Johnson: "Can't get 24 bit color to work with PGX64"
- Previous message: dave dickerson: "Re: VXFS vs. UFS w logging"
- In reply to: Darren Dunham: "Re: why the difference in swap value for vmstat output"
- Next in thread: Darren Dunham: "Re: why the difference in swap value for vmstat output"
- Reply: Darren Dunham: "Re: why the difference in swap value for vmstat output"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: 13 Nov 2003 20:52:49 -0800
thanx. I know I was confused :-(.
I like to confirm my theory, given that scanner starts scanning when
free memory is lower than lotsfree, the "free memory" is really RAM,
hence even if I add swap, it would not help the scanner, right ?
Darren Dunham <ddunham@redwood.taos.com> wrote in message news:<ISUsb.10080$MR.8517@newssvr29.news.prodigy.com>...
> In comp.unix.solaris Bill <bill12317@lycos.com> wrote:
> > All,
>
> > I am looking into a performance issue in one of our servers and need
> > some help to understand the output of vmstat and top as I am now
> > totally confused.
>
> > The machine is a Sun Enterprise Server running Solaris 8, with 40G of
> > RAM and 56 G of disk swap
>
> WOW! 56G of disk swap? What are you doing with this box that makes
> that useful? Normally if that much stuff is swapped out, you've got
> performance issues.
>
> , which means I have about 90 G of swap.
> > Checking the output of vmstat -p 5, I have the following, I
> > understand, the swap + free - RAM used by kernel, but I expect
>
> > a. swap + free will be close to 90G, but 57 + 3 is close to 60G, can
> > anyone explain why does my swap go ?
>
> The first is free swap (VM) and should agree with 'available' in swap -s
> output. The second is free RAM. Free RAM counts in both columns.
>
> > b. I understand the swap or free will be changed, but I explect to
> > total of swap + free be the same, but why did it go from 60 G (57 + 3)
> > to 70 G ( 62 + 8) ?
>
> 5GB of RAM was freed for some reason.
>
> > c. I used to think that only high number for sr means a memory
> > shortage,
>
> On Solaris 8 and higher, that's generally correct. Actually *any* sr
> would point to that.
>
> but
> > what if my system is a database server using a lof of file system, and
> > given we have high fpi, will we be benefited by increasing more disk
> > swap, so as to cache more file system ?
>
> Filesystem pages are cached in free RAM, they're not swapped to disk so
> increasing your disk backing store should have no effect on file system
> caching.
>
> > The following is the top output ...
>
> > ===============================================================
> > Memory: 36G real, 7199M free, 33G swap in use, 60G swap free
> > ==============================================================
>
> > From what I understand from the man page,
> > therefore the above it means that out of the 36G of RAM, about 29G had
> > been used, right ? because it shows 7 G is free.
>
> Looks right to me.
>
> > does it also mean that 33G of disks swap is being used? because
>
> > "swap in use" is the amount of swap area on disk that is being used
> > "swap free" is the amount of swap area on disk that is still
> > available.
>
> If you use 'swap -l' to measure swap on disk in use, it is more clear.
>
> > Does the 33G swap include physical RAM ?
>
> Probably. Compare it to 'swap -s' which does include physical RAM.
>
> > I thought I understand swap here, but after looking at the data for a
> > few days, I am confused. thanx for the help
>
> 'swap -s' looks only at the unified VM space and tells you allocation
> within it.
>
> 'swap -l' looks only at disk swapfiles and tells you how much is used.
- Next message: Dave Johnson: "Can't get 24 bit color to work with PGX64"
- Previous message: dave dickerson: "Re: VXFS vs. UFS w logging"
- In reply to: Darren Dunham: "Re: why the difference in swap value for vmstat output"
- Next in thread: Darren Dunham: "Re: why the difference in swap value for vmstat output"
- Reply: Darren Dunham: "Re: why the difference in swap value for vmstat output"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]