I was running Nagios 1.2 + nagiosgraph on a FC3 box and had a similar problem. The system would run for about a week and then the CPU utilization would saturate and the system became unresponsive. I started monitoring the box more closely and what I found was that the nagios process was slowly using more and more memory. Once all the memory was used up it would start thrashing the swap space hard.
I used Nagios and nagiosgraph to chart the CPU utilization and it would go up logrithmically once available memory became low so the system would appear to be fine until the last bit of memory was used up and then the CPU would reach staturation very rapidly.
To temporarily fix the problem I would restart the nagios process every four days or so. When the nagios process was restarted it would release all the memory it was holding and start over.
I'm currently testing nagios 2.0b4 on a FC4 box and I no longer see the problem. I'm not sure if it was something in FC4 or nagios 1.2.
The two issues sound similar, however, yours seems to occur much faster and there is a big difference in hardware. I was running nagios on a 600MHz Via Epia system with only 128MB of RAM. That xSeries eServer has much more power and likely more RAM. Since you experience the problem much faster you could do somethying like 'vmstat -a 30 > vmstat.txt' for the 7 hour period and if it is memory you'll have a log of the memory and swap usage.
Edited Sun Nov 27 2005, 11:11AM ]