Memory Problemas on HPUX 11.31

[CentralM]

Hi all,
We have HPUX 11.31 on SD1,Oracle Retail 13.1. The memory is on 96% used and the process batch can´t run.
UNIX95=1 ps -e -o vsz,pid,ppid,args | sort -rn | head -20
336192 5041 1 ora_dbw1_RIBPP
336192 5039 1 ora_dbw0_RIBPP
336000 5025 1 ora_pmon_RIBPP
335920 5055 1 ora_d000_RIBPP
332544 5051 1 ora_mmon_RIBPP
332288 17053 17052 oracleRIBPP (DESCRIPTION=(LOCAL=YES)(ADDRESS=(PROTOCOL=beq)))
332288 10585 1 oracleRIBPP (LOCAL=NO)
332288 10583 1 oracleRIBPP (LOCAL=NO)
332288 10550 1 oracleRIBPP (LOCAL=NO)
332288 10503 1 oracleRIBPP (LOCAL=NO)
332288 10494 1 oracleRIBPP (LOCAL=NO)
332288 10484 1 oracleRIBPP (LOCAL=NO)
332288 10472 1 oracleRIBPP (LOCAL=NO)
332288 10455 1 oracleRIBPP (LOCAL=NO)
332288 10453 1 oracleRIBPP (LOCAL=NO)
332288 10444 1 oracleRIBPP (LOCAL=NO)
332288 10438 1 oracleRIBPP (LOCAL=NO)
332288 10414 1 oracleRIBPP (LOCAL=NO)
332288 10406 1 oracleRIBPP (LOCAL=NO)
332288 10404 1 oracleRIBPP (LOCAL=NO)
cmuxxxx01:/>swapinfo
Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 8388608 1633960 6754648 19% 0 - 1 /dev/vg00/lvol2
reserve - 6754648 -6754648
memory 9925952 5627488 4298464 57%

max_acct_file_size 2560000 Default Immed
max_async_ports 4096 Default Immed
max_mem_window 0 Default Immed
max_thread_proc 3001 3001 Immed
maxdsiz 2063835136 2063835136 Immed
maxdsiz_64bit 4294967296 Default Immed
maxfiles 2048 2048
maxfiles_lim 65536 65536 Immed
maxrsessiz 8388608 Default
maxrsessiz_64bit 8388608 Default
maxssiz 134217728 134217728 Immed
maxssiz_64bit 1073741824 1073741824 Immed
maxtsiz 100663296 Default Immed
maxtsiz_64bit 1073741824 Default Immed
maxuprc 3687 3687 Immed
Please if you can help me..
Regards
C.

[Dennis Handly]

>The memory is on 96% used and the process batch can´t run.

Were there any particular error messages?
Please provide the "swapinfo -tam" output. Was the one you provided during this 96% used phase?

You could add more swap or RAM.

[CentralM]

Hi,

We can add swap and mem, but the problem is that are 7 vpars in SuperDome1, they have memory on 96% and some 100%.

cmuxxxx01:/>swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 8192 1596 6596 19% 0 - 1 /dev/vg00/lvol2
reserve - 6596 -6596
memory 9693 5382 4311 56%
total 17885 13574 4311 76% - 0 -
cmuxdbpp01:/>

[Dennis Handly]

>We can add swap and mem, but the problem is that are 7 vpars in SuperDome1, they have memory on 96% and some 100%.

Why do you say "but"? If you can add swap and memory, then do it. Otherwise you will have to limit the applications you run.

dev 8192 1596 6596 19%
memory 9693 5382 4311 56%
total 17885 13574 4311 76%

Currently you are only using 76% of swap.
What do you have configured for the filecache?

[CentralM]

I say "but", because they are 7 Vpars and I need know if we can resolv this problem without add memory, only tuning it. If is mandatory add memory I will do it only in the critical servers, because the free memory there isn´s enough for all servers.

filecache_max 5082087424 Default Auto
filecache_min 508207104 Default Auto

Regards.
CM

[James R. Ferguson]

HI:

> filecache_max 5082087424 Default Auto

With default settings, your 'filecache_max' is 50% of your memory and your 'filecache_min' is 5%.

I would try reducing the upper limit to about 10% or 15% and measure I/O performance with 'glance'. Use 'kctune':

# kctune filecache_min=5% filecache_max=10%

The 11.31 filecache supports both memory-mapped files via 'mmap(2)' and file I/O using the standard 'read(2)' and 'write(2)' calls. This differs from previous releases where the buffer cache was used only by 'read' and 'write'.

Regards!

...JRF...

[Dennis Handly]

>because they are 7 Vpars and I need know if we can resolve this problem without add memory, only tuning it.

Now it's clearer.
Is it much easier to add swap?

[CentralM]

yes, I am going do two things:

1.- change the values filecache_max =10%
and filecache_min=5% like say James R.

2.- Add swap like you say and check how the servers will work.

Thanks
CM.