Help CPU pegged at 96% but cannot find culprit

JB mmyymm at hotmail.com
Wed Jan 11 21:52:32 UTC 2006


T is nt good for a computer.. But I needed 1 week to figure-it out ;) . Mine
was to freeze every 1 hour.. At that time I had about 2 millions apache hits
+ mysql heavy. So I woke up (Nagios sent me an sms and rebooted server at
nght)
Due to this load decreases, I apply php eaccelerator and loade was steady to
about 40.. 
Then I changed the kernel in grub.conf and everything went back to normal
like a bad dream. Now that server has 3 millions hits per day and load is
around 3, proc load between 20-40%.
This bug is reproducible and I can prove with my cacti graphs. 

The difference between a bad kernel and a good one is on how much swap will
show.. a bad one will never show more then 50 megs or so while a good one
will cache almost everything. 

JB
 


-----Original Message-----
From: redhat-sysadmin-list-bounces at redhat.com
[mailto:redhat-sysadmin-list-bounces at redhat.com] On Behalf Of Romeo
Theriault
Sent: Wednesday, January 11, 2006 7:02 PM
To: redhat-sysadmin-list at redhat.com
Subject: Re: Help CPU pegged at 96% but cannot find culprit

I have about 18 instances of apache showing up with a `ps -aux |  
more`   . I have a ton of coldfusion instances though. I just found  
that I don't have all of the latest updates to coldfusion, which I  
will apply tonight and see how it deals with it then. I'll write back  
to the list to tell you if that fixed it , but how would I go about  
getting a lower version of the kernel to downgrade if II needed too?

Also, is it harmful for the computers cpu to be this high for a day  
until I can work on it at night?    95% cpu usage.

Thanks,

Romeo

On Jan 11, 2006, at 11:34 AM, JB wrote:

> How many apache processes do u have?
> I had the same issue and the culprit was kernel ES  
> (2.6.9-22.ELsmp). I am
> using an old kernel 2.6.9-11.E on all my RH servers. Based on my  
> cacti it
> looks like the new kernel has a different and bad way of using swap.
> Top indicated 400 Apache processes with little each ;) Now I have  
> the same #
> of processes but that amount is slightly different.
> I tried with a newer kernel and was the same thing. It happen as  
> well on a
> system without HT.
>
> JB
>
> -----Original Message-----
> From: redhat-sysadmin-list-bounces at redhat.com
> [mailto:redhat-sysadmin-list-bounces at redhat.com] On Behalf Of Romeo
> Theriault
> Sent: Wednesday, January 11, 2006 3:27 PM
> To: redhat-sysadmin-list at redhat.com
> Subject: Help CPU pegged at 96% but cannot find culprit
>
> Fresh install of red hat enterprise AS 4 update 2 with all the
> patches, running as a web server(apache) with coldfusion. The cpu is
> pegged at around 96% with 65 or so under user and 30 or so for the
> system. I've done a top and nothing is showing as using this much
> cpu. I downloaded ibm's performance tool nmon and tried it and it's
> showing the same stats. How can I find the process which is taking up
> so much cpu. Thank you very much.
>
> Romeo Theriault
>
> --
> redhat-sysadmin-list mailing list
> redhat-sysadmin-list at redhat.com
> https://www.redhat.com/mailman/listinfo/redhat-sysadmin-list
>
> --
> redhat-sysadmin-list mailing list
> redhat-sysadmin-list at redhat.com
> https://www.redhat.com/mailman/listinfo/redhat-sysadmin-list

--
redhat-sysadmin-list mailing list
redhat-sysadmin-list at redhat.com
https://www.redhat.com/mailman/listinfo/redhat-sysadmin-list




More information about the redhat-sysadmin-list mailing list