Hi All,

I am doing core reduction on RHEL 6 servers, during testing phase what is observed is after core reduction , putty session stops.
ping to server also fails.

however server is up, (logged directly on windows hypervisor).

I ran following script at boot ,(creating a sym link in /etc/rc3.d directory)

#!/bin/bash

for i in {12..15}
do
echo "Reducing core cpu$i"
echo 0 > /sys/devices/system/cpu/cpu$i/online
echo "Core Reduced (cpu$i)"
done
echo "$(nproc) cores available"

##################################

server remains up for sometime then putty session closes

From logs i can see
: CPU 15 is now offline
Oct 26 15:56:41 NWLOG kdump: kexec: unloaded kdump kernel
Oct 26 15:56:41 NWLOG kdump: stopped
Oct 26 15:56:41 NWLOG kdump: kexec: loaded kdump kernel
Oct 26 15:56:41 NWLOG kdump: started up
Oct 26 15:56:41 NWLOG kdump: kexec: unloaded kdump kernel
Oct 26 15:56:41 NWLOG kdump: stopped
Oct 26 15:56:42 NWLOG kdump: kexec: loaded kdump kernel
Oct 26 15:56:42 NWLOG kdump: started up
Oct 26 15:56:42 NWLOG kdump: kexec: unloaded kdump kernel
Oct 26 15:56:42 NWLOG kdump: stopped
Oct 26 15:56:42 NWLOG kdump: kexec: loaded kdump kernel
Oct 26 15:56:42 NWLOG kdump: started up
Oct 26 15:56:42 NWLOG kdump: kexec: unloaded kdump kernel
Oct 26 15:56:42 NWLOG kdump: stopped
Oct 26 15:56:42 NWLOG kdump: kexec: loaded kdump kernel
Oct 26 15:56:42 NWLOG kdump: started up
Oct 26 15:56:42 NWLOG kdump: kexec: unloaded kdump kernel
Oct 26 15:56:42 NWLOG kdump: stopped
Oct 26 15:56:50 NWLOG irqbalance: WARNING, didn't collect load info for all cpus
, balancing is broken,,
for all cpus i made offline.

Please suggest.

Thanks

You're using RHEL6....the best suggestion would be to CALL RED HAT SUPPORT....you are PAYING FOR RHEL, RIGHT??? They can assist you with an sosreport and analyze things. Otherwise, you don't tell us what kind of VM this is, on what kind of hardware, or what exactly you're trying to accomplish. The message seems fairly clear...you've set the system up with a certain number of CPU's, then (essentially) yanked one out, so things aren't 'balanced' anymore. Reboot the server with the new CPU settings, and see what happens.

Since this is for 'testing', you have done something that causes your system to crash. So...DON'T DO IT, and consider the test over.

Well, thats the thing, if We had support from redhat i wouldnot be posting this here. Seems logical isnt it ? Anyways if u have solution please suggest else dont bother replying on the post. We have to do the core reduction activity thats why asking at this open forum to seek some expert advise. My organisation didnt take support from redhat. Valuable suggestions are welcome.

Read the LQ Rules about text-speak and not using it. And if you are NOT paying for RHEL, then why bother using it, when CentOS is totally free, and identical. Seems logical, doesn't it??? You use RHEL for stability, extensive testing, and SUPPORT FOR WHEN YOU HAVE ISSUES. Otherwise, there are LOTS of free distros you can use, that can do everything RHEL does.

And if you want to get snotty about "dont bother replying" when someone asks you questions or points out the VERY obvious "call support for the product you should be paying for" solution, don't bother POSTING.
..and just saying 'we have to do the core reduction activity', doesn't answer the questions asked. Your 'organization' needs to pay for the commercial products it uses, or use the free products.

AGAIN: what are you trying to accomplish and why??? The 'valuable suggestions' would be for you to answer the questions you've been asked, and think about what's been said. Again...you aren't providing details, and the error seems very clear. Since you aren't paying for RHEL, you obviously don't have the irqbalance bug fix that was reported in the RHEL knowledgebase, you can't even download it, or read the customer-only errata report.

So: pay for RHEL or use CentOS...those are your two fixes. Pick one.

1 members found this post helpful.

Wonder if the hardware is at fault/not supported. I'm going on this error too. "rqbalance: WARNING, didn't collect load info for all cpus
, balancing is broken,, "

Look up some of the other web pages on that fault maybe??

1 members found this post helpful.

I've successfully used that technique in the past on RHEL6 - but never in a tight loop like that. Try adding a sleep to the script to allow the offline action (moving running tasks to another CPU) to complete before offlining the next.
It's possible tasks are being moved to (say) 13 and then you try to offline 13 immediately.
Easy to test - I'd use a decent value, say 30 seconds.

1 members found this post helpful.

I doubted that may be the issue, the tight loop and also the irqbalance Warninig. Can you suggest a good Advanced BAsh Scripting book, I have read many, but didnt find that quality in any of the ones availabe on internet

Thanks for your time

I have always used the Advanced Bash-Scripting Guide available here. Download as appropriate.