Momentary "Pausing" of my system.

[Vijayabaskar_3]

OS:HP UX B 11.11
Server: RP5430/9000 L33

My system "pauses" or "hangs" momentarily for a couple of seconds (10-15 seconds) frequently. Because of this, I am not able to use it either at my application level or at my command prompt level.

After that the system behaves normally.
But my users think that the system hangs and they are struk.

Any suggestion to resolve this is most welcome.

Thanks,
Vijayabaskar.

[Thierry Poels_1]

Hi,

do you have enough free memory available? System "pauses" could be caused by heavily swapping.
(sar -w to monitor swapping)

regards,
Thierry.

[Jean-Louis Phelix]

Hi,

This could also be related to a very high ninode (not used anymore with vxfs), or a very big buffer cache (dbc_max_pct too high). This can 'hang' the system while cleaning.

Regards.

[Vijayabaskar_3]

Hi,
I tried sar -w to monitor the usage,
but my system says,

sar: Can't open /var/adm/sa/sa26

I do not have a sa folder.

New bee: How do I activate SAR?

[Bernhard Mueller]

Hi,

run e.g.
sar -w 1 10
to run it 10 times in 1 second intervals

Regards,
Bernhard

[Vijayabaskar_3]

Hi,
Please find the attached find,with the
sar -w 1 10 report.

B.Regds,
Vijayabaskar.

[Jean-Louis Phelix]

hi,

you can also enable system activity collect to be able to run sar on longer periods of time. You just have to enable it from crontab and have some free space available. Useful during performance problem ... See 'man sadc'.

Regards.

[john korterman]

Hi,
can you rule out a network problem? A simple non-scientific test is to ask the users to - when they are experiencing the problem - activate some other application requiring network connection and check if that application is also affected.

regards,
John K.

[Vijayabaskar_3]

Hi John,
I can rule-out the network issue as it also happened to me on a weekend when I was in the office alone with no users logged into the network.

Regds.

[Chris Wilshaw]

To enable sar;

As root

Make the directory:
mkdir /var/adm/sa

Edit the root cron file:
crontab -e

Add the line below, and save the file:
0 * * * * /usr/lib/sa/sa1 1200 3

This will automate collection of sar stats at 20 minute intervals

[john korterman]

Hi again Vijayabaskar,
it may not necessarily be your application that affects the network; it could very well be something else. But if you can rule out the network as the source, then try and open a window and run "top" continuosly in that window. When the problem appears then look at the cpu usage in top. It could be that some process used all the cpu power of the machine for a while. Again, not very scientific, but it is difficult to telle what to look for. Other ideas would be to check for messages appearing in the /var/adm/syslog/syslog.log and the apllication log during the problematic perios and see if they can give a clue.

regards,
John K.

[Graham Cameron_1]

When you were alone at the weekend were you directly connected at the console or other hardwired terminal? If not, it may still be a network problem.

Continuing the non-scientific theme, I would be tempted to fire a background job from the console:
--
while sleep 1
do
date >> /tmp/datelog
done &
--
Then, after your next hang, check the datelog file.
Is there a gap corresponding to your outage?
If yes, call HP; if no, system is fine, it's your access to it that is the problem.

-- Graham

[Mike Stroyan]

These three patches have fixes for momentary hangs. Check that your system has them or superseding patches
===================================================================
PHKL_28695 s700_800 11.11 Cumulative VM, Psets, Preemption, PRM, MRG
PHKL_28410 s700_800 11.11 vm preemption point, pdc, vhand performance
PHKL_25233 s700_800 11.11 select(2) and poll(2) hang

[Vijayabaskar_3]

Hi Mike,
I think, I dont have those patches. I may need to install them.

Looking at my SAR collection, I find that the load average is 2.52, 3.14, 3.43
and 78% of swap is being used.
I think I may have to update the patches.

When I work on the console, I do not face that much of a problem.

Regards

[john korterman]

Hi again,
if you do not face much of a problem when working on the console, I think it indicates that the major part of the problem lies in the surroundings. Try Graham's suggestion and check what is going on in your lan during the problematic periods.

regards from the non-scientific
John K.