number of mails in /var/mail can increase load?

[Pavel Moravec]

Hi all,
I am investigating why system loads of HP 11.11 boxes at our customer are permanently increasing while the boxes process the same amount of work. I figured out that more affected are the boxes that have more emails under /var/emails (in general, the number of such emails is 10000 on average).

Please, can somebody confirm or disagree that there is a dependency between high number of emails and system load? Or is it a random coincidence?

Some data from the boxes:
top shows:
CPU LOAD USER NICE SYS IDLE BLOCK SWAIT INTR SSYS
avg 1.89 45.0% 0.0% 18.5% 36.5% 0.0% 0.0% 0.0% 0.0%
vmstat runqueue length is 6 in average (two processors there)
No disk bottleneck, no network bottleneck, the only problem is high CPU usage.

Thanks and regards,
Pavel Moravec

[Kapil Jha]

Hello Pavel,
there is nothing i can disagree on what you said.mails can create some high CPU utilization.
Can you please let us know which process is consuming most of the CPU???
is it sendmail/postfix??

U have to see which process is culprit.

BR,
Kapil+

[sreekanthtm]

Hi,

Huge no: of mails will not affect the system performance. But you can have a look in those mails, weather its pointing to any bad cron tab entries.

>>>No disk bottleneck, no network bottleneck, the only problem is high CPU usage.

Use top command to get information about the processes, mainly CPU utilization. Compare the TOP o/p with other server (Which is working fine).
Is the h/w configuration of all the server are same?? Check the memory and swap utilization also.

# swapinfo
# glance

Rgds
Sreekanth

[Pavel Moravec]

Thanks for your replies.

# swapinfo
Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 4194304 959524 3234780 23% 0 - 1 /dev/vg00/lvol2
reserve - 2283892 -2283892
memory 3071232 1171088 1900144 38%

Neither swap or memory seems to be the bottleneck. Glance isn't installed there.

The highest load is caused by a Java application that serves as a router of SMS messages. We are also examinating if the load can't be caused by the application itself only but since the application has been restarted several times (due to other reasons) and the load is still increasing, we don't think so that the application is the only cause.

All boxes have the same HW, patches etc., and should have very similar work to perform.

Yes, most of the emails were sent by crontab, I have fixed this but have not permission to delete the emails yet (some administrative reason only).

[Kapil Jha]

Hey Pavel,
If ur mails are very old and they have been stuck because of some reason , to check you can move the directory smwhere and see how ur system is behaving.If it is the real cause of slowness you can go ahead and take the decision if u want to keep the mails and delete them , and if that not the reason we can look at other side of the issue.

Hope this help,
BR,
Kapil+

[sreekanthtm]

Hi,

From the swapinfo, System is having only 2GB swap space and its reserved. Please increase the swaparea. You have 3GB physical memory installed (used 2GB). At least you should have 6GB swap.

Are you using any JAVA related appl. If not you can kill those processes and reduce the CPU utilization.

Rgds
Sreekanth

[Steven E. Protter]

Shalom,

Lots of mail in the outbound send queue will increase system load.

http://www.hpux.ws/?p=6

SEP

[TTr]

> that there is a dependency between high number of emails and system load?
There could be an indirect dependency. How are your users reading the emails? Locally (elm, pine, mailx) or remotely (pop, imap)?

> TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
> dev 4194304 959524 3234780 23% 0 - 1 /dev/vg00/lvol2

Your server has done 959MB of disk swapping. If this is active swapping it could be the source of your problem. Adding a little physical memory to the server would help. Or look if you can conserve memory usage by finding and stopping any unecessary processes that are running.

[Aneesh Mohan]

Hi Pavel,

To check the largest memory eating process

UNIX95=1 ps -e -o vsz,uid,pid,ruser,args |sort -rn |head -n 20


Aneesh

[Suraj K Sankari]

Hi,

I think there is no connection between /var/mail and CPU load

So if your CPU load is more then check which process is taking more CPU

Suraj