HPUX 11.0 memory not freed

[tecnomatix]

Hi All,
I am using c3600 machine with 512MB memory and 1500MB swapspace.
I am trying to install our application and the memory is getting preaty full and preaty quickly.
after killing the the installation the memory doesn't free itself.
there is nothing new in ipcs -ma...
Thanks,
Ayelet

[Stefan Farrelly]

What makes you think memory isnt being freed up after you killed it ? use the vmstat and the free column to verify how much free memory you currently have (number is in pages so multiply by 4096).

If memory still isnt being freed up did you kill it nicely (-15) so it had a chance to tidy up before dying or kill -9 ? (always try a normal kill first before a -9).

It could easily be a poorly written application that doesnt free up all the memory it uses when it terminates or dies in which case only a reboot will fix it (or get the application developers to look into it and fix it).

[T G Manikandan]

Also how much lockable memory you have.
Might be that your kernel parameters are high and the lockable memory is very less on the machine.
you can use

UNIX95= ps -e -o ruser,pid,vsz=Kbytes|more
to check the individual process--memory usage values.

THanks

[tecnomatix]

Thanks alot for your quick respond,
This installation is working great on c200 (with the same amount of memory)
i kill the process with ^C because otherwise it stucks the machine.
i see the memory is not freed with:
panther:/opt/tmp [root] > swapinfo -mta
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 1536 0 1536 0% 0 - 1 /dev/vg00/lvol2
reserve - 315 -315
memory 376 214 162 57%
total 1912 529 1383 28% - 0 -

this is my kernel parameters:
dbc_max_pct 30
dbc_min_pct 5
max_async_ports 50
max_fcp_reqs 512
max_mem_window 0
max_thread_proc 256
maxdsiz 2063835136
maxdsiz_64bit 0X0000000040000000
maxfiles 200
maxfiles_lim 1024
maxqueuetime 0
maxssiz 83570688
maxssiz_64bit 83570688
maxswapchunks 768
maxtsiz 0X40000000
maxtsiz_64bit 0X0000000040000000
maxuprc 75
maxusers 32

[tecnomatix]

anyone has other suggestions?

[Mike Stroyan]

It would be more meaningful to look at before and after views of swapinfo -mta. We can't tell what kind of growth in virtual memory use the application caused.
If the swapinfo total remained the same, then a transfer from reserved to memory categories really doesn't mean any additional resources were consumed.
A process can't just leak memory after it exits. It can create shared memory objects that persist after exit. You would have seen that in ipcs output.
A process can also make requests of other processes like the X server or swagentd that cause them to grow their sizes.
It is easy to write a program using XCloseDownMode that leaks big X server pixmaps.
That is not a common defect.

[Jeff Schussele]

Hi Ayelet,

Having dbc_max_pct at 30% means you could be using up to 150 Mb + for disk buffer cache. This may be much more than you need & costing you application memory space.
You could try backing it down to the 10 - 15 range & gain some obviously needed memory space.

Rgds,
Jeff