Can anyone beat this? I saw this on a server today.

you are the winner!!! :D

how long did it take you to get that info :)
Surprised that the server did not crash!

cheers,
:beer:

what does it mean? The server load time?

Whoaaa and I thought we were in deep trouble when I saw our load went up to 70!!

Originally posted by sweethorizons
what does it mean? The server load time?

Yikes! Them there is a smokin' server...

Those are the load averages for the server...

load average: 937.39, 626.00, 112.49

937 = current, 626 = 5 minute average, 112 = 15 minute average

1.00 would mean 100% processor utilization and most people would probably keep it in the 3.00-8.00 range for the most part.

Mine stays around the 0.00 for all three most days...

load of 1.00 does not mean 100% CPU utilization ... then load of 2.00 will mean 200% CPU utilization? Can't be right, because by definition, the highest the CPU can be utlized at any moment is fully utilized, i.e. 100% utillized :)

cheers,
:beer:

Well u kinda got it right. It actually goes 1min 5min 15min

the first section is 1min and correct me if i'm wrong about these but it mean that for 1 minute a process was running for this percent of time. so if it was 0.60 that would be 60% of 1 minute a process was running, if it's over 1.00 then basically it's just rolling over into the 5 min section and so on.

The load average numbers give the average number of jobs in the run queue over the last 1, 5, and 15 minutes. (These three time periods may vary from one vendor's Unix system to another, but are usually 1, 5 and 15 minutes.) In other words, the n-minute load average is the number of processes competing for the attention of the CPU(s) at any moment, averaged over n minutes.


From:
http://lithos.gat.com/docview/load_average.html

Cheers
:beer:

No more 90000% CPU utilization ok :D

thx muppie ... I was looking for that link :)

cheers,
:beer:

dam the highest ever i got on my server was 267+ when i have a memory leak but 937+ :eek:

eva2000,

how do you detect a memory leak and what programs caused it? How do you solve it? Lemme know ...

cheers,
:beer:

Originally posted by dektong
eva2000,

how do you detect a memory leak and what programs caused it? How do you solve it? Lemme know ...

cheers,
:beer: it was caused by an earlier installation of APC PHP Cache as posted here http://vbulletin.com/forum/showthread.php?s=&threadid=14184&highlight=cache

i have latest version and not memory leaks :D

thx eva ... will read now :)

cheers,
:beer:

Originally posted by ShaunM
Well u kinda got it right.

It seems I kinda got most of it wrong actually :D DOH!


The load average numbers give the average number of jobs in the run queue over the last 1, 5, and 15 minutes. (These three time periods may vary from one vendor's Unix system to another, but are usually 1, 5 and 15 minutes.) In other words, the n-minute load average is the number of processes competing for the attention of the CPU(s) at any moment, averaged over n minutes.


I always thought it was a percentage thing, but regardless, so this means that a load average of 0.02 (let's say this is on a 15 minute average) would mean there are 0.02 processes competing for the attention of the CPU? Iimplying that 98 more could run before *any* were competing? And 1.00 means there is one process competing for the attention of the CPU (would this mean that the CPU was 100% utilized?) Again, if this were a 15 minute average, thus I mean an average of 1... (fill in rest...)

A load average of 0.02 means that there is an average of 0.02 processes in the run queue. This might mean that 98% of the time the processor is doing nothing and the remaining time it has a single process to run; it might mean that 99% of the time the processor is doing nothing and 1% of the time there are two processes doing stuff; it might mean that the processor is idle 99.9% of the time and the remaining 0.1% of the time there are 20 processes doing stuff.

The cpu utilization will always be lower than the load average, but the difference between them depends upon how often two differing processes simultaneously want cpu time.

Originally posted by eva2000
it was caused by an earlier installation of APC PHP Cache as posted here http://vbulletin.com/forum/showthread.php?s=&threadid=14184&highlight=cache

i have latest version and not memory leaks :D


Hmmm I read the APC docs they are still warning people about potential leak on the mm version.... which one are you using, eva2000 ?

I am thinking of using another cache and avoid APC :D

Originally posted by muppie



Hmmm I read the APC docs they are still warning people about potential leak on the mm version.... which one are you using, eva2000 ?

I am thinking of using another cache and avoid APC :D using shm since mmap doesn't seem to want to work on my server :(

What's confusing me is that I heard of the execute and ready state of the processes. At a certain time only one process can be in the execute state. (one processor machine). The other processes which are ready for execution wait in the ready state or queue.

What's now the run-queue. Are these all processes in the ready and execute state? Or only those in the ready state?

nòóx

I don't know exactly the story of load average, but from your description:

A processor basically can only execute one thread/process at a time. In single user operating system such as DOS, this is the case. Say you run chkdsk, then chkdsk is the only program that is running and using the CPU.

In a multi tasking operating system like unix, there can be many processes running at the same time. The cpu runs all of them one by one by switching from one process to another. The amount of time one process gets served by the CPU is called cpu time slice.

To the processes themselves, it appears that they are running simultaneously because they don't need to reliquish the CPU to another processes, they simply don't have to care.

This method of multi tasking is called pre-emptive multi tasking, because the CPU itself pre-empt the processing and moves to another process etc.

yeah. I know that. I've studied computer science. :D

Maybe my English is not that good but with "At a certain time only one process can be in the execute state." I meant that each process is only executed for a short time slice.

It's just the term "run queue". If these are only those processes in the ready queue or additionally the one which is currently executed in the CPU (if there is one to execute).

In relation to the load average, as was stated, is the average number of processes that are *waiting* to be served... hence in your term.. those that are in the "run-queue" I think

Silly people. The load average is computed by the VM manager; the VM manager can only do this when it is running; thus when the load average is computed there are no processes running on a single-processor machine.

On a multiprocessor machine the load average measures the average number of waiting and executing processes.

Originally posted by cperciva
Silly people. The load average is computed by the VM manager; the VM manager can only do this when it is running; thus when the load average is computed there are no processes running on a single-processor machine.


Why not? Does the VM manager not know how long a process was executed? Why should it be different on a single and a multiprocessor machine. Doesn't make sense.

The VM manager counts all processes currently waiting or running. When the VM manager is running, the maximum number of processes running = #procs - 1.

The scary part is that I was able to telnet in and still function! Eventually, when it reached that level, while trying to run uptime it just spewed out all kinds of garbage. Im sure the load average was probably higher then 937.39, but that was the highest I was able to read. And the most amazing thing was that I was able to reboot the server!



Oh in case your wondering..The server specs are 1.4GHZ CPU, 512RAM with the new RH 7.2. We were testing the stablilty of RH 7.2 with a few scripts I made, and tried to simulate the worst case scenerio . Held up pretty good eh?

Originally posted by LinuxGroup
Oh in case your wondering..The server specs are 1.4GHZ CPU, 512RAM with the new RH 7.2. We were testing the stablilty of RH 7.2 with a few scripts I made, and tried to simulate the worst case scenerio . Held up pretty good eh?


Would you care to post those scripts, there might be one or two people that would like them. And would you define what you consider is the Worst case scenerio.

thanks.

Sqposter / Michael

Sorry, I dont feel it safe or fair to release these scripts. The reason being is that these scripts do not require root access to execute and could cause alota unneeded issues with webhosting COs.

Originally posted by LinuxGroup
Sorry, I dont feel it safe or fair to release these scripts. The reason being is that these scripts do not require root access to execute and could cause alota unneeded issues with webhosting COs.

Ahhh, Dos attack type scripts (assuming), Well understood that release into the wild these types of scripts could cause much havac.

here take a look at this report http://grc.com/dos/grcdos.htm , he had a attack on his server from a script kiddy, fought it and won.

I got to admit, I never thought of one of those scripts as a decent way of testing a server peak.

Sqposter / Michael

Actually yesterday I jumped on one of our servers because it had a 3 min downtime (bastards) and saw the 5 min load average was 300. I am sure we got pretty high up there.. No clue who did it but we got a baseball bat waiting for the guy. :)

Originally posted by sqposter


Ahhh, Dos attack type scripts (assuming), Well understood that release into the wild these types of scripts could cause much havac.

here take a look at this report http://grc.com/dos/grcdos.htm , he had a attack on his server from a script kiddy, fought it and won.

I got to admit, I never thought of one of those scripts as a decent way of testing a server peak.

Sqposter / Michael

Actually, it was an on server script. I would not call it a DoS attack. Im supprised this script has not been created before and used to crash servers.

And yes, these scripts are great ways to test server peak. I do more than create scripts to stress the server, such as creating scripts/programs that will monitor and stomp such processes without using much load in the script/program itself.

some time before i was reading about intel's forthcomming processor itenium and current xeon proccessors and it was something like these proccessors can execute more then one proccess in single clock cycle. but os should be optimized for full utililzation of such proccessor.
:)

Actually justin was testing it on my machine and I though it was funny until it caught on fire and destroyed my whole computer. I was mad about having to replace my computer which is farely new but was even more mad because now I have to replace a house.

I just wish I had insurance.

Originally posted by LinuxGroup
Actually, it was an on server script. I would not call it a DoS attack. Im supprised this script has not been created before and used to crash servers.
There are many scripts like that. In fact I'd say less then three months ago a poster on Bugtraq said PHP could be used for such an attack. Eventually another person corrected his post to show that you can perform such attacks with anything right down to good old DOS .bat files. Simple loops are all it takes to make the CPU die under heavy pressure. I bet most programmers have accidentally created a loop in their software at one point or another and upon executing it you better believe they noticed. :)

Regarding what Graham said, the CPU does get incredibly hot under non-stop usage. If it wasn't properly cooled I can see it overheating but the new P4's (from what I was told) automatically decrease their speed as they over heat. Another words you can't fry one because it will just slow down to the point where it doesn't need cooling. Haven't tested this in person but I do recall someone telling me this.

Anyway, unless you had the case off and paper against the CPU or something else that could start on fire, I fail to see how it could have burned your house down. Also, unless you bought the home outright chances are your mortgage company would require home insurance. In fact it would be required before closing and if you couldn't prove insurance you wouldn't get the title or they would insure it through their companies preferred partner for a higher rate.

ROFL!!:D

actually it was just a joke man. ROFLMAO!:D

Originally posted by Chicken


1.00 would mean 100% processor utilization and most people would probably keep it in the 3.00-8.00 range for the most part.

Mine stays around the 0.00 for all three most days...


Actually I think a load number at 1 means that on average over the time period specified, there has been 1 process waiting to run. I'm not sure that necessarily translates to 100% CPU usage. For instance over 15 minutes, there could have been 1 minute with 15 processes waiting and 14 minutes with 0 waiting and thus a 1 load, but only 1 minute out of 15 was there 100% cpu usage.

Now 937 load is pretty sick. I cannot imagine what the server was doing to get that high. I am impressed that it didn't thrash itself to a kernel panic...

Take care,

Brian

Originally posted by graham
actually it was just a joke man. ROFLMAO!:D
I knew it sounded strange and was from a person with just one post so I wasn't sure what to make of it. :)

Originally posted by graham
Actually justin was testing it on my machine and I though it was funny until it caught on fire and destroyed my whole computer. I was mad about having to replace my computer which is farely new but was even more mad because now I have to replace a house.

I just wish I had insurance.

:eek: :eek3: :eek2: