UNIX for Dummies Questions & Answers

Hi

what does %idle reflect in SAR command

suppose SAR commamd is showing idle as 90%

does it means sytem is fine or sytem is in danger state.


below is the o/p of sar command and %idle is 90 then what does its means.

Code:

 
sar 1 4
AIX ab41cp01 3 5 000B3E0AD400    01/01/15
System configuration: lcpu=8  mode=Capped 
10:13:20    %usr    %sys    %wio   %idle   physc
10:13:21      52       7       0      92    4.00
10:13:22      52       7       0      91    4.00
10:13:23      57       7       0      86    4.00
10:13:24      52       7       0      91    4.00
Average       53       7       0      90    4.00

It means your system is bored. It is looking for something to do about 90% of the time?

Hi Don

does it means it means that server is not working properly ??
i executed the above cmd on server.

higher value is good or lower value of %idle is good .

---------- Post updated at 12:09 PM ---------- Previous update was at 12:07 PM ----------

i was checking server overload through sar command.

i am not sure if higher value is good or lower .

Strange,

%usr %sys %wait and %idle do not add up to 100..

Anyone seen this before ?

@scriptor: are you sure this output is right?

If you want to have hardware configured to get the job done at minimum costs, you would expect low %idle. If you want a system that is always available and will respond immediately to requests, you would expect high %idle. Is 10am on New Year's day a good time to determine the load on your servers? Did you have a representative load on your servers at that time?

Since I have no idea how you are trying to configure your server, no idea whether these four seconds represent "normal" use of your server, and (as Scrutinizer said) no idea why your numbers don't add up; I have absolutely no idea whether low %idle is good or high %idle is good.

This is the output of an AIX LPAR (logical partition, a kind of virtual machine) and therefore the explanations before - while being correct for physical machines - are not really true.

The IBMs p-Series (this is the sole hardware to support AIX) hypervisor dynamically adjusts the layout of the system depending on its own load, the load of other LPARs on the same "managed system" (the IBM-speak for "hardware node", physical system) and the LPARs profile parameters, therefore the single presentation of some sar-output without any accompanying information is next to useless.

First off, the OS version seems to be 5.3, which is out of support for nearly 2 years now. I suggest to update ASAP.

Post an output of vmstat -tw 1 while the system is under a typical load. Notice that "%usr", "%sys", "%idle" and "%wait" are only relative to "ecc" (entitled capacity comsumed). For example a 50% usr with an ecc of 0.1 means that the system is exerting effectively 5% of the processor ressources it could muster on user processes. It further means the system has 10% of its processor possible resources allocated but the hypervisor doesn't see the need to give it more right now.

We will also need to know which hardware your managed system is (Power5, Power6, Power6+, Power7, Power7+ or Power8), the LPAR profile you booted from and let us see from there. You might need to post additional information before any verdict is possible.

You might want to read this thread in the meantime for a short explanation on analyzing vmstat-output.

I hope this helps.

bakunin

So are you saying that %usr %sys %wait and %idle do not add up to 100.. in an LPAR? That does not seem logical to me. I can understand that the actual CPU power being used is reflected in the physc metric. But using that relative measure the numbers still should add up to a 100, I would think, no?