FC-1 died mysterious several times a week

[Jeff Vian]

Andrew Robinson wrote:

> Guolin Cheng wrote:
>
>>
>>  It is quite strange that my FC-1 Box died mysteriously several times 
>> a week. now it died again. I can ping it from other host, and nmap it 
>> without issues, but when I try to ssh to it, or access it by means of 
>> nfs, I got no response.
>>
>>  The machine box is a HP workstation xw4100, 2x4G CPU, 4G memory, 
>> 4x250GB SATA drives(two connected by means of Intel ICH5 on 
>> mothboard, 2 by means of Promise RAID S150 TX4), the kernel is 
>> upgraded 2.4.22-1-ntpl-2174smp.
>>
>>  I can not find errors in /var/log/messages. I have been trying to 
>> install a model FC-1 machine to replace our existing RH8.0/9.0 since 
>> the latter has no suuport soon. sigh...
>>
>>  any suggestions are greatly appreciated.
>>
>
> I have had similar problemss. It turned out that I was filling up my 
> root  partition. I initially configured my system with one partition 
> for swap and one for a monolithic filesystem. I moved /var and 
> /usr/share to partitions with more space, and my system stopped freezing.
>
> I never actually caught the system in the act of filling up the root 
> filesystem. I deduced the problem from available clues and experience 
> with other *nixes. Like you, there were no errors in 
> /var/log/messages. The "df" command showed my root filesystem usage in 
> the mid-90 percent range. Apparently after I rebooted the system, 
> something was cleared out of /var or /tmp to bring the utilization 
> back below 100%.
>
This is why it is often recommended that filesystems with dynamic usage 
(/home, /tmp, /var, /usr/local, etc) should have their own partitions.   
If they fill up it does not halt the entire system and since root is not 
in the /home filesystem you can still log in and take action to fix the 
problem without having to reboot.

Monolithic filesystems are BAD for several reasons, and this is one. 
(upgrade problems is another)

> I don't know if this is your problem, but it might be worth checking out.
>
> HTH,
>
> Andrew Robinson
>
>