HP-UX Itanium Swap Question

[Nyck_1]

In the syslog file on one of my Itanium servers I'm getting rather a few of the following messages:-

vmunix: Deferred swap reservation failure pid:

When I run a swapinfo -tam I get the following output:-

swapinfo -tam
Mb Mb Mb PCT START/ Mb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 16384 204 16180 1% 0 - 1 /dev/vg00/lvol2
reserve - 15454 -15454
memory 32739 24284 8455 74%
total 49123 39942 9181 81% - 0 -

So from this I need to add some more device swap?

If this machine was having swap issues would not the /dev/vg00/lvol2 be reading 100%? From what I can see its just using 1% of swap

[Ivan Krastev]

Find the program with reported PID and use chatr to see if the lazy swap is enabled.

chatr program | grep lazy

Solution is to disable it.


regards,
ivan

[Amit Parui]

Hi Nyck,

Seems your swap is not used up. Check out for any shared memory issues. Also check whether the existing swap is configured properly. Lastly, look for increasing the memory.
---------
Else refer to the following links.

forums.itrc.hp.com/service/forums/questionanswer.do?threadId=94295

http://docs.hp.com/en/5187-4198/ch02s09.html

I hope you get some idea.

[Nyck_1]

It looks like this issue is caused by some rogue oracle prcoess on this server:-

ps -fl |awk '{print $1,$2,$3,$4,$5,$10,$15}'
F S UID PID PPID SZ COMD
1401 S oracle 12639 1 30592 oraclebbtv_lon
1401 S oracle 10402 1 30728 ora_pmon_bbtv_lon
1401 S oracle 10404 1 31820 ora_dbw0_bbtv_lon
1401 S oracle 10406 1 32844 ora_lgwr_bbtv_lon

The SZ for these process is 30728, how do I work out what this is in MB?

[Dennis Handly · Accepted solution]

>The SZ for these process is 30728, how do I work out what this is in MB?

ps(1) is your friend, it says it is 4Kb.
So it is about 120 Mb.

[Nyck_1]

This is also verified with the below:-

kctune |grep -i page
pagezero_daemon_enabled 1 Default Immed
volpagemod_max_memsz 6144 Default Immed
vps_pagesize 4 Default

i.e vps_pagesize 4

[Geoff Wild]

I wonder why it is only 4?

On my Itanium systems it is 16:

# kctune |grep page
pagezero_daemon_enabled 1 Default Immed
volpagemod_max_memsz 6144 Default Immed
vps_pagesize 16 Default Immed


Good news - it is dynamic.

What does kcusage show?

Mine:

# kcusage
Tunable Usage / Setting
=============================================
filecache_max 8162492416 / 8166457344
maxdsiz 54919168 / 2063835136
maxdsiz_64bit 9043968 / 4294967296
maxfiles_lim 128 / 4096
maxssiz 229376 / 134217728
maxssiz_64bit 786432 / 268435456
maxtsiz 696320 / 100663296
maxtsiz_64bit 201326592 / 1073741824
maxuprc 60 / 3780
max_thread_proc 36 / 3000
msgmni 2 / 4200
msgtql 0 / 4200
nflocks 146 / 4096
ninode 1256 / 8192
nkthread 466 / 8416
nproc 231 / 4200
npty 0 / 60
nstrpty 6 / 60
nstrtel 2 / 60
nswapdev 1 / 32
nswapfs 0 / 32
semmni 32 / 4200
semmns 491 / 8400
shmmax 4664078336 / 17179869184
shmmni 9 / 512
shmseg 2 / 300


Rgds...Geoff

[Nyck_1]

kcusage
Tunable Usage / Setting
=============================================
dbc_max_pct 14 / 50
maxdsiz 58540032 / 1073741824
maxdsiz_64bit 174489600 / 4294967296
maxfiles_lim 209 / 4096
maxssiz 1114112 / 134217728
maxssiz_64bit 1179648 / 1073741824
maxtsiz 147456 / 100663296
maxtsiz_64bit 53248 / 1073741824
maxuprc 928 / 4090
max_thread_proc 69 / 256
maxvgs 4 / 10
msgmni 1449 / 6546
msgseg 0 / 32767
msgtql 0 / 6596
nfile 11975 / 65536
nflocks 26 / 4096
ninode 7855 / 34816
nkthread 1888 / 8416
nproc 1242 / 4200
npty 0 / 60
nstrpty 14 / 60
nstrtel 29 / 60
nswapdev 2 / 10
nswapfs 0 / 10
semmni 67 / 4096
semmns 16894 / 32767
shmmax 268435456 / 4294967296
shmmni 43 / 512
shmseg 3 / 300

You mention that yours is 16, did you specify this at build time or did you change it to be 16 at a later date?

[Nyck_1]

You mentioned the following:-
>The SZ for these process is 30728, how do I work out what this is in MB?

>ps(1) is your friend, it says it is 4Kb.
>So it is about 120 Mb.

So to come to the 120 Mb total itr would just be the case of doing the following:-

4 X 30728 = 122912 which is 120MB?

[Dennis Handly]

>Geoff: I wonder why it is only 4?

Because you don't want the ps(1) output to be different depending on the VM page size. Hopefully that matches top(1).

>So to come to the 120 Mb total it would just be the case of doing the following:
>4 X 30728 = 122912 which is 120MB?

Well, 4*30 = 120 is close enough for government work. ;-)