Glance and swapinfo mismatch

[Matt Hearn]

Hi all! I have a customer who knows just enough about HP-UX to be dangerous, and he's been playing around in Glance, which he says indicates that memory and swap usage are very high. I'd have to agree, based on what I'm seeing here:

Mem Util 93% 93% 93%
Swap Util 68% 68% 68%

The fact that those values haven't changed at all makes me a little worried about how accurate glance really is. Glance -m shows the following:

Page Faults 0 34025 0.0 195.0 2754.2
Page In 0 9817 0.0 56.2 787.1
Page Out 0 0 0.0 0.0 0.0
KB Paged In 0kb 0kb 0.0 0.0 0.0
KB Paged Out 0kb 0kb 0.0 0.0 0.0
Reactivations 0 0 0.0 0.0 0.0
Deactivations 0 0 0.0 0.0 0.0
KB Deactivated 0kb 0kb 0.0 0.0 0.0
VM Reads 0 0 0.0 0.0 0.0
VM Writes 0 14 0.0 0.0 2.6

That doesn't seem to indicate a great deal of swap usage. If I check swapinfo -a I get:

Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 4194304 0 4194304 0% 0 - 1 /dev/vg00/lvol2
dev 12288000 0 12288000 0% 0 - 0 /dev/vg00/lvol9
reserve - 16482304 -16482304
memory 13456488 3965760 9490728 29%

Those are TOTALLY different values from what Glance is reporting, and would seem to be well within norms. Which one is lying? I hear that I might need a patch to make Glance and swapinfo match up, but we're patched to the "equivalent of December 2004" (our customer gets customized patch bundles from HP). We should be pretty close to up-to-date. I've been reading ftp://eh:spear9@hprc.external.hp.com/memory.htm, but that deals more with figuring out what's chewing up memory, and not with why swapinfo and glance are at odds.

Thank you!!

[Kent Ostby]

Matt --

Read this web page first to get an idea of what swapinfo is reporting:

ftp://hprc.external.hp.com/memory.htm

Essentially the memory line reported by swapinfo is the pseudo space line so it is completely different than what Glance is showing.

[Matt Hearn]

Ah, I missed the stuff at the end about how to understand what I'm seeing. Now it all makes more sense.

Right now it looks like I have 16GB of memory, and about 1.2GB is free. So I shouldn't expect to have any actual performance problems, right? The amount of free memory isn't low enough that paging should be occurring, correct?

The customer is just concerned about the high usage of memory, but if I can convince him that this is normal and doesn't impact his application, he'll go away.

Thanks!

[Simon Hargrave]

If your buffer cache has not been tuned away from the default then chances are most of your memory is being used a buffer cache.

What is the output of:

kmtune | grep -E "dbc|bufpages"

on this server?

[Pedro Cirne]

Hi,

It's a normal situation to have Memory Usage around 93%.

You should only be concerned if you start to have page out/deactivatoins activity, and they all are ZERO.

So...it's just normal behavior.

Enjoy :)

Pedro

[Gavin Clarke]

I tend to look for paging out as my first indication of memory pressure:

Here's a link that covers looking at vmstat to check for paging out.

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=691648

I hope you don't mind me throwing another tool into this mix.

[Gavin Clarke]

Oh and high memory use can be a good thing, it means your not paying for something that you aren't using.

I note that glance also reports no page outs at all, which looks good to me.

[Matt Hearn]

Hey Simon:

root@********> kmtune | grep -E "dbc|bufpages"
bufpages 0 - (NBUF*2)
dbc_max_pct 5 - 5
dbc_min_pct 2 - 2

Thanks!

[Cem Tugrul · Accepted solution]

Hi,
your swapinfo -tam seems normal
i also agree with Pedro

Good Luck,

[Cem Tugrul]

also your
dbc_max_pct 5 - 5
dbc_min_pct 2 - 2
parameters good for me