Craziest Problem I ever seen!
 

Hello, I just got a new dedicated, AMD 2800 with 1 gig ram, 120 gig HD. Now I installed a few rpms, like php-mysql and a bunch of others, manually installed some PLesk 7 files and it went good. But now when I check the processes with the "top" command I see this:

PU states: cpu user nice system irq softirq iowait idle
total 0.0% 0.0% 0.0% 33.3% 32.9% 33.7% 0.0%
Mem: 968560k av, 85664k used, 882896k free, 0k shrd, 7128k buff
26352k active, 45260k inactive
Swap: 2104504k av, 0k used, 2104504k free 49820k cached

Whats up with the irq/softirq/iowait each at 33% ?? Before I touched it the load was idle at 99.9% now its idle 0% :(

Anyone have an idea how to fix it ? I rebooted 3 times and still get this...

Welcome to LQ!

That's an interesting question. What kernel version are you running (show us the results of uname -r, if you could) and is it one you custom compiled? Can you also try running "vmstat 1 15" and seeing what the results look like, in particular the number of interrupts handled.

I skim over the Linux kernel mailing list and I think I recall someone reporting something like this, but it was with a 2.6 SMP kernel IIRC. I'm not sure if your situation is even exactly the same, so a few more details would be helpful.

Hi, thanks for taking the time to help :) My system is running Fedora core 1 and I didnt copmiled anything, I just installed rpms basically. I tried the uname -r command, got this: 2.4.25

I also tried the vmstat, here is the result:

[root@host root]# vmstat 1 15
procs memory swap io system cpu
r b swpd free buff cache si so bi bo in cs us sy wa id
1 0 0 884736 6328 48960 0 0 79 16 111 30 0 1 0 98
0 0 0 884728 6328 48960 0 0 0 0 107 9 0 0 0 100
0 0 0 884728 6328 48960 0 0 0 0 103 12 0 0 0 100
0 0 0 884728 6328 48960 0 0 0 0 104 11 0 0 0 100
0 0 0 884728 6328 48960 0 0 0 0 106 12 0 0 0 100
0 0 0 884720 6336 48960 0 0 0 12 109 16 0 0 0 100
0 0 0 884720 6336 48960 0 0 0 0 106 12 0 0 0 100
0 0 0 884720 6336 48960 0 0 0 0 105 10 0 0 0 100
0 0 0 884720 6336 48960 0 0 0 0 104 14 0 0 0 100
0 0 0 884720 6336 48960 0 0 0 0 107 9 0 0 0 100
0 0 0 884720 6336 48960 0 0 0 0 106 14 0 0 0 100
0 0 0 884720 6336 48960 0 0 0 0 104 9 0 0 0 100
0 0 0 884720 6336 48960 0 0 0 0 104 12 0 0 0 100
0 0 0 884640 6344 48960 0 0 0 20 113 26 0 0 0 100
0 0 0 884712 6344 48960 0 0 0 0 112 17 0 0 0 100

This is a remote system, thats why I have much difficulty fixing it, I would like to deinstall a shitload of stuff to see if it makes any difference, ddunno how :(

Also when checking the "top" I see that the buffer size is increasing all the time in 4k incrementation,and never go down always up.. I worry it may crash the system if I let it run for long.. Here is the top again I let it run a bit, you see the huge buffer number at the right of screen :(

61 processes: 60 sleeping, 1 running, 0 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait idle
total 0.0% 0.0% 0.0% 33.3% 32.9% 33.7% 0.0%
Mem: 968560k av, 84120k used, 884440k free, 0k shrd, 6488k buff <-------- Getting bigger!
26308k active, 43800k inactive
Swap: 2104504k av, 0k used, 2104504k free 48960k cached

Thanks again for your help :)

It's normal for the Linux kernel to allocate as much memory as possible for buffers. After all, if the memory isn't being used for something, it's just going to waste. The kernel will give that memory back to the application if it needs to. Looking at your vmstat it seems that most of the CPU time is devoted to the idle loop, which is what you'd expect. Can you look at the processes listed in top and see if ksoftirqd (or any other "k" process) are taking inordinate amounts of CPU time. Otherwise, I'm not sure if there's anything really to be concerned about here.

BTW -- I've never seen a top with that exact header you've listed. Unfortunately, I don't have a FC1 system to look at this more closely with. Have you applied all the errata to your FC1 install?

Hello my friend, I havent installed myself FC1, it was installed from the company I got the box. I got a clean box and the system was idle at 99.9% when I got it and did a top command. Now its at 0% Idle thats what blows :( Theres no processes that takes more than 0.2% CPU, just in the header the damn irq/softirq/iowait that goes 33% each for 2-3 secs, then it goes back to 0% for 2-3 secs and so on.. I cant find anyone that had that problem :(

If you look at the last column of your vmstat output (the one under the 'id' heading) that's the percent of time spent idle in the last report period. In your case it's 99 or 100. What's strange is that the output of top that you showed us does not agree with this vmstat output. In any case, I figured it out -- it's a bug in Fedora's top, see http://bugzilla.redhat.com/bugzilla/...buglist=111216 for more details.

YOUHOOO !!! Fix the damn box guys !! It was a rpm of Plesk fucking with the IO, I uninstalled it and then it stopped the imap pop3 and then idle is back at 100% finally, no more spikes of iowait/irq/irqsoft :)

Thanks for trying to help :)

Lol thats wicked, I shutdown -r my system and then some processes that was supposedly deleted came back.. and now the error is back in top wtf, this is wicked.. Is there any file that is like the config.sys or autoexec.bat like in windows where it says which program to start on startup ?

Thanks

Another update, I manually stopped the imap server and the load idle is back to 100% Seems the IMAP fucks the top or something or is misconfigured and looking for something, how can I conffigure my IMAP ?

Thanks !