topic Re: monitoring nflock , system table in Operating System - HP-UX

monitoring nflock , system table

DeafFrog — Mon, 11 May 2009 09:49:56 GMT

Dear Gurus ,

We had some "EDEADLK Deadlock condition" with one of application S/W and the vendor advised use to monitor the
nflock nfile system resources.
1)when nflock is full will not an error message be logged in syslog ?
2)we don't have wma/extract , we have glance , can we use glance to collect statistics over a period of time ?
(glance -t >> ~xyz/report is not read friendly )
3)from where can i download Krmond to enable the kernel resource monitor, sorry, but i am not able to find the same on software.hp.com.OS is B.11.00 .

Re: monitoring nflock , system table

Ganesan R — Mon, 11 May 2009 10:13:12 GMT

Hi,

Do you mean nflocks?

To prevent file or data corruption, when simultaneous access to a given file by two or more processes, a process that is altering the contents of a file must be able to lock the file against conflicting uses until it is safe to release control of the file. This is what nflocks is used.

>>1)when nflock is full will not an error message be logged in syslog ?<<

Yes. you will get a message like below in system when nflocks is full.
cannot lock xxxxxxxx: No locks available

>>2)we don't have wma/extract , we have glance , can we use glance to collect statistics over a period of time ? <<

I don't think so. You need that to extract the data. Normally lsof will be used to find the nflocks usage.

Re: monitoring nflock , system table

Duncan Edmonstone — Mon, 11 May 2009 10:25:45 GMT

assuming you have a recent enough version of glance, just do the following:

echo 'print GBL_STATTIME, " ", TBL_FILE_LOCK_UTIL, "%"' > /tmp/advisor.cfg

glance -aos /tmp/advisor.cfg -j 60

This will give you an indication of the utilisation level of nflock every 60 seconds.

I'm suyre you can figure out the rest from that

HTH

Duncan

Re: monitoring nflock , system table

DeafFrog — Mon, 11 May 2009 10:44:39 GMT

Thanks Ganesan and Duncan.
Glance version is C.03.71.00.The /tmp/advisor.cfg file contains nothing expect headder.

Re: monitoring nflock , system table

Duncan Edmonstone — Mon, 11 May 2009 10:54:55 GMT

Hmmm I didn't think I'd need to spell it out...

echo 'print GBL_STATTIME, " ", TBL_FILE_LOCK_UTIL, "%"' > /tmp/advisor.cfg

glance -aos /tmp/advisor.cfg -j 60

The file /tmp/advisor.cfg is a configuration file to tell it what data to monitor - it's not an output file. Glance with write ou
tput to stdout unless you redirect it. so you could instead do:

glance -aos /tmp/advisor.cfg -j 60 > /tmp/nflocks.out

you should get an entry in /tmp/nflocks.out every 60 seconds

HTH

Duncan

Re: monitoring nflock , system table

DeafFrog — Mon, 11 May 2009 11:25:50 GMT

Dear Duncan ,

/tmp/nflocks.out will not be readable , as i mentioned in the initation post.

Re: monitoring nflock , system table

Duncan Edmonstone — Mon, 11 May 2009 11:31:19 GMT

so when you run:

glance -aos /tmp/advisor.cfg -j 2

what *do* you see on the screen... anything at all?

HTH

Duncan

Re: monitoring nflock , system table

DeafFrog — Mon, 11 May 2009 11:50:36 GMT

Dear Duncan ,

From what i can see on screen, i cannot make out how to infer the nflocks value.
It's something relating to process Id that did the last context switch, i guess .
.Attachment . My idea was to collect nflocks value in a file over a period of time , in a system which has only glance.

Re: monitoring nflock , system table

Duncan Edmonstone — Mon, 11 May 2009 12:08:22 GMT

hmm - I wonder if the "-aos" option isn't working on your older version of glance. On my version (04.50) I get this:

# glance -aos /tmp/advisor.cfg -j 2

Welcome to GlancePlus

14:01:10 0.2%
14:01:11 0.2%
14:01:13 0.2%
14:01:15 0.2%
14:01:17 0.2%

which is showing me only 0.2% usage of my file locks...

Maybe try this syntax instead:

# glance -advisor_only -syntax /tmp/advisor.cfg -j 2

HTH

Duncan