System slowness with latest updates

[Steven Haigh]

On 12/03/2006, at 10:18 AM, David Timms wrote:
> dragoran wrote:
>> Steven Haigh wrote:
>>> On Sat, March 11, 2006 3:17 pm, Steven Haigh wrote:
>>> As a followup to this, I've noticed that the syslog is around 3-4  
>>> minutes
>>> behind the events that happen. say I login to an IMAP server, the  
>>> maillog
>>> shows the login at 16:57:48 - if I breat out of that and show the  
>>> date,
>>> the clock will show 3-4 minutes later (ie 17:01) - even though  
>>> the line
>>> has only appeared seconds ago.
>>>
>>> Even things such as spamassassin is taking 217 seconds to scan a 2Kb
>>> email. snmpd results to localhost are timing out - even though  
>>> the process
>>> is listening.
>>>
>>> It takes 45-50 seconds to log out of a su shell back to the  
>>> logged in
>>> user, the same for becoming root via 'su -'.
>>>
>>> I have never seen this kind of behaviour on a system before -  
>>> especially
>>> when the CPU is 99% idle. Has anyone come across this type of  
>>> behaviour
>>> before?
>>>
>>> the system is:
>>> model name      : AMD Sempron(tm) Processor 2800+
>>> stepping        : 0
>>> cpu MHz         : 1602.495
>>>
>>> It has 1 Gb of RAM, and seemed to work perfectly until last  
>>> nights updates.
>>>
>>>
>> have you tryed a different kernel?
> And which one are you using while seeing this ?
>
> Can you check /etc/hosts and or post it here; I saw similar things  
> when I accidently removed my machine host name from /etc/hosts ? Or  
> maybe it doesn't have the localhost entry and net stuff is timing out.
> ping localhost
> ping {machine host name}
> ping {gateway}
>
> Second guess, that I have noted with win machines at the office -  
> every now and then win will set the netcard link to 100/Full or  
> alternatively to Auto, when it needs to be the other way (for win  
> to boot properly - depends on net card). This makes everything on  
> the machine run really slowly (opening any app - which tries to  
> reference network printers, because the card then communicates at  
> 10/half, but in short responses  - and never gets more than 100  
> Bytes/sec etc). I have never seen this on Fedora, though.

Thanks for the suggestions! I'd tried most of these before I posted.  
The network cards I have *always* negotiate the wrong duplex setting  
with my cisco switch and I have to lock them at 100/FD with mii-tool.  
This wasn't the issue however.

After playing around more and more and having things sometimes work,  
sometimes not, I thought it may be a hardware fault - so I swapped  
over a lot of hardware to see if that was the case (maybe dud ram?).  
This made no difference. I even went as far as reinstalling the  
entire system with FC4 and applying all updates - with no change.

What I finally found out is that as I use this box for remote logging  
(cisco equipment logs to it), it seems that the -x option was not in  
the config file (/etc/sysconfig/syslog) which was causing the system  
to look up EVERY entry for EVERY machine that tried to write a line  
to the syslog. As soon as I added -x to the syslog command line and  
restarted syslog, the system started purring away as it has done for  
the last few years.

I wish the -x option would be there by default - as it would have  
certainly saved me a LOT of pain. :)

--
Steven Haigh

Email: netwiz at crc.id.au
Web: http://www.crc.id.au
Phone: (03) 9017 0597 - 0412 935 897