top output

[Sanjay Tailor]

Hello,

I have a D370 on 10.20. Usually, when I run top the LOAD column of the CPU States is very low, about 0.4 to 0.5. But today, it is averaging 1.2 to 1.5. There are no large processes running, no abnormal sessions etc. I have checked every option of sar and cannot find anything wrong. I have included some of the output I get. Where and how is the LOAD column calculated. I also looked on the LCD display on the box itself and it too is averaging F11F to F21F when it is usually F01F. What could the system be doing. Is there a way to find out. ps and sar and top seem to be unable to tell me what maybe happening.

Any ideas? Thank yo for your help.

Load averages: 1.35, 1.65, 1.67
184 processes: 183 sleeping, 1 running
Cpu states:
LOAD USER NICE SYS IDLE BLOCK 1.35 2.2% 0.0% 0.8% 97.0% 0.0%

Memory: 51564K (9600K) real, 58868K (18476K) virtual, 24980K free Page# 1/14

Sanjay.

[Brian M. Fisher]

Is the system I/O bound?
Try "sar -d 10 10" and look at the results.

Brian
<*(((>< er

[Sanjay Tailor]

Hello,
This is the output I get from sar -d 10 10:

09:05:27 device %busy avque r+w/s blks/s avwait avserv
09:05:37 c0t5d0 3.80 0.50 4 45 4.04 11.70
c0t8d0 1.60 0.50 2 26 3.81 28.70

verage c0t5d0 3.88 0.75 4 45 6.33 14.12
Average c0t8d0 2.49 18.72 3 49 134.51 39.10
#

Although it seems as if c0t8d0 is waiting a long time, at this time of the day there are not many users logged onto the system. In fact I had only 12 users logged on and I was getting the same type of output. My system will average 75 users and I still won't get output from top like this. Could I possibly be having a hardware disk problem? Or something else?

Thanks for your help.
Sanjay.

[Sanjay Tailor]

Hello,
This is the output I get from sar -d 10 10:

09:05:27 device %busy avque r+w/s blks/s avwait avserv
09:05:37 c0t5d0 3.80 0.50 4 45 4.04 11.70
c0t8d0 1.60 0.50 2 26 3.81 28.70

verage c0t5d0 3.88 0.75 4 45 6.33 14.12
Average c0t8d0 2.49 18.72 3 49 134.51 39.10
#

Although it seems as if c0t8d0 is waiting a long time, at this time of the day there are not many users logged onto the system. In fact I had only 12 users logged on and I was getting the same type of output. My system will average 75 users and I still won't get output from top like this. Could I possibly be having a hardware disk problem? Or something else?

Thanks for your help.
Sanjay.

[CHRIS_ANORUO]

run ps -ef|grep prosses name (eg oracle)|wc -l

[Sanjay Tailor]

Hello,

ps gives me no insight. Right now I have about 50 users on with 200 processes. This is normal. There seems to be no abnormal or unusual processes running. Could one of them run away? How can I find out?

Thanks,
Sanjay

[CHRIS_ANORUO]

The GLANCE product will allow you to pick a process, and then see the open files on that specific process, as well as memory useage etc.

You should have a TRIAL version of Glance on your OS cd's.
Or you can down "lsof" from cs.utah.edu site.
Install and run from /opt/lsof/bin/lsof

[John_Hancock]

This could be due to a hung process. These can be tricky to find. If there is a process in the process list find the parent and kill it. Otherwise it will be a process that is not responding. I had this with the snmp daemon. It would not respond to snmp requests and the load average was flat-linig at one.

One indication of a hung process is a runq flat-lining at one. You can use xload to show a graph of runq.

John Hancock