Description of problem: HP-OVPA(Open view performance agent) is a tool used to monitor the system performance. The daemon Scopeux is the main component which collects system data and logs into the log file. This component gets started at the boottime by default or can be manually started after boottime. Now the problem is the size of Physical memory got from the /proc/meminfo file. Basically Scopeux takes the data from /proc/meminfo and logs it into the logfile. When Scopeux is started as part of system bootup, the value seen in measurement for physical memory is over 100 megabytes larger than the value seen when scopeux is restarted once bootup is completed. Showing physical memory swinging up and down by 100 megabytes at a time when there is no physical change to the system is a problem. Version-Release number of selected component (if applicable): How reproducible: System Details: uname -a Linux ovrlxt66.rose.hp.com 2.6.18-8.el5 #1 SMP Fri Jan 26 14:16:09 EST 2007 ia64 vi /etc/issue Red Hat Enterprise Linux Server release 5 (Tikanga) Kernel \r on an \m Steps to Reproduce: 1. Note the Physical memory(MemTotal) got from /proc/meminfo. 2. reboot the system 3. Note the Physical memory got at boottime. You can see a difference in Physical memory got at boottime to what you got after boottime. Actual results: At boottime MemTotal: 981.0 MBytes After boottime MemTotal: 863.0 MBytes Expected results: At boottime MemTotal: 863.0 MBytes After boottime MemTotal: 863.0 MBytes Additional info:
Hm... I thought to get know the first real user of procinfo utility and it's only a misfit bug. This is not for procinfo -- it's for the kernel.
Why is this a problem? Just becuase a system doesn't change physical state, doesnt mean we can't change the amount of ram present during the boot process, or after the boot process for that matter. xen dom0 kernels do this to provide system memory to guest kernels. Some arches remove the zero pages from the totalrampage count during boot up since those pages will never be used. It may look a bit funny in your monitoring, but there isn't anything wrong with it.