Defunc services (Unwanted services)

[Omprakash_2]

Hi,

Here i am handling HP UX B.11.00 U 9000/785, I want to know how to find out the unwanted services running background. can any one provide me the command for the same.

Note: In AIX the command will be #ps -ef | grep defunc

[Ganesan R]

Hi Omprakash,

What do you mean by unwanted services?. Services is different and "defunc processes" are something different.

You can use the same command in hp-ux also to list the defunc processes.

[Trng]

Hi,

ps -ef | grep defun will do and if you found multipel zombie process then u have to take action...

[Dennis Handly]

If you want to take action against zombies, you need to use the advanced search and look for zombie master. And also look for the one with "wanted poster". :-)
http://forums.itrc.hp.com/service/forums/questionanswer.do?threadId=1278104

[Omprakash_2]

Hi,

Kindly find the attached screen shot, the process java & vxfsd is running for a long time. How to stop these service, can you help me to know the Impact.

If multiple users are connected, automatically my server gets very slow. I have checked the home size its normal like 39%, but when multiple users connected means its raised upto 90%. Hence users will not able to access the database in the server.

How to list the zombic process in HP ux servers??? ps -ef | grep defunc, displays nothing.

[Dennis Handly · Accepted solution]

>the process java & vxfsd is running for a long time. How to stop these service, can you help me to know the Impact?

You don't stop root processes because they have been running for a long time. You may stop them because they are taking 100% of the CPU. You shouldn't stop vxfsd.
To figure out what those java processes are doing, you need to get the process tree:
UNIX95=EXTENDED_PS ps -xHfu root

>If multiple users are connected, automatically my server gets very slow.

Showing the output of top and "swapinfo -tam" during these times would be helpful. You can use "top -d 1 -n 30 -f top.output" to get the top output into a file, so you don't need a screen shot.

>I have checked the home size its normal like 39%, but when multiple users connected means its raised upto 90%.

Home size? What's that, the bdf output?

>How to list the zombie process in HP-UX servers? ps -ef | grep defunc, displays nothing.

Then you don't have any and they don't take any resources, except a process slot.

[Omprakash_2]

Hi,

Thanks for your valuable informations. You asked about home size its nothing but the home directory for all users.

How to check the Hardisk in servers (bad sectors), without rebooting the server.

[Avinash20]

Refer

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=543493

[Omprakash_2]

Hi,

Excatly my issue is We are using MFGPRO ERP Application. Every day we are loading Data files to Database through CIM program, Normally it ill take 6 hours to load. But Nowadays its taking nearly 10 hours. For reducing the time , Kindly suggest us to monitor the server performance (Harddisk. Memory, LV size)

[Suraj K Sankari]

Hi,

Data is coping where to where?
Local to local or local to remote?

In this case you have to monitor your network, disk as well as your server performances.

Suraj

[Omprakash_2]

Hi,

Absoultly you are right, data coping in the same server only.

Can you provide me procedure to trouble shoot this issue. Specially i am asking about hardisk errors, how to check the installed harddisk is having error or not

[Aneesh Mohan]

Hi Omprakash

1) For checking physical volumes this may help you...

for i in `strings /etc/lvmtab |grep "dsk"`
do
echo "Now checking Physical Volume $i "
pvdisplay -v $i |grep -e stale -e unavilable |wc -l
done


2) For checking the resources bottle you can use sar,iostat,vmstat commands

just an example

sar -u 5 10

If %usr is high more than 70% then we can say the system is likely experiencing a CPU bottleneck due to user process

If %wio have value more than 15 ,then it sounds like disks are involved in the bottle neck

sar -d 5 10


|Aneesh
If %busy for any disks is greater than let say 50 and the same disk is avwait > avserv then it seems to be an I/O bottle neck on the device

[Omprakash_2]

Hi,

Can you explain the mentioned command.

#ddif=/dev/rdsk/c3t3d0 of=/dev/null bs=1024k.

You have mentioned some scripts, but i am not able to understand the scripts.

My first target is i need to check Harddisk performance. If possible explain me in briefly step by step...

[Ganesan R]

Hi Omprakash,

Why do you suspect harddisk badsectors for the delay on data copying? You need to dig into system to find out the actual cause.

If any badsectors in the harddisk, system will through read write I/O error in the syslog.
If you still want to check harddisk for media errors you can use this command.

#dd if=/dev/rdsk/cxdxtx of=/dev/null bs=1024k

I would suspect your data volume size could have increased suddenly that makes more time to copy. You need to compare with old data volume size.

There are many things you can try like distributing the data into multiple disks by striping the destination volume, routing the i/o to multiple paths etc..

Check the resources performance using glance when you do copying..

[Ganesan R]

>>>Can you explain the mentioned command.

#ddif=/dev/rdsk/c3t3d0 of=/dev/null bs=1024k. <<<

There should a space inbetween dd and if like below.

#dd if=/dev/rdsk/c3t3d0 of=/dev/null bs=1024k

This command will read the data block by block(block size is 1024k=1MB) from the input device /dev/rdsk/c3t3d0 and send it to output device /dev/null(where the data will be nullified)

If any badsectors during read operation dd will through read I/O error. If no bad sectors, in records and out records will be same at the end of the command execution.

[Avinash20]

I have just checked your snapshot you attached dated " Feb 14, 2009 04:46:30 GMT"

Only 90 MB free memory available on the server
This is too less, and might be the reason for the Slow backup

Could you stop the application and check if you are able to take the backup.

Also check
# vmstat 2 5

Check for po(page-outs)
This should not take place and should not be double digit.
If its double digit, then there is a serious memory pressure on the server which needs to be address
You could get kmeminfo script from HP, or some other tools (eg: glance) to check which proceses are utilizing lot of physical memory on the server

Best of Luck !!