Where is all my memory going?

I recently rebuilt a Pentium D 820, 2gb DDR2 server and fitted it with OpenSUSE 11.0 x64. It was previously running fine under 10.1 x64, using a very similar loadout.

The problem is that this time around, this server is SLOW. The slowness acts a lot like its constantly loading things out of swap. The server is a personal, headless machine of mine that primarily hosts a pair of VMWare server virtual machines and the Azureus BitTorrent client.

The two virtual machines are:
a Windows Server 2003 server running MS Exchange (1024mb RAM)
an Ubuntu virtual appliance running Restore-Backup (256mb RAM)

According to VMware, both those machines are currently using about 163 megabytes total.

Other major programs running in the background include Apache, nfsd, samba, webmin, sshd, as well as one instance of X and vncserver. Suse’s firewall is disabled in favor of shorewall.

This is my output from top:


top - 22:36:22 up 8 days,  2:27,  3 users,  load average: 5.39, 3.00, 1.55
Tasks: 150 total,   1 running, 149 sleeping,   0 stopped,   0 zombie
Cpu(s):  1.3%us,  2.8%sy,  0.0%ni, 30.6%id, 64.7%wa,  0.0%hi,  0.7%si,  0.0%st
Mem:   2059548k total,  2043864k used,    15684k free,     1640k buffers
Swap:  3148700k total,  1646276k used,  1502424k free,   685728k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
25868 tom       20   0 1506m 605m 3992 D    0 30.1 186:29.76 kio_sysinfo
 2818 root      10 -10 1211m 213m 207m S    2 10.6 441:59.18 vmware-vmx
 3907 tom       20   0 1097m 161m  12m S    2  8.0 113:50.19 java.bin
 4268 tom       20   0  690m 150m 3072 D    1  7.5  30:09.70 Xvnc
 7199 tom       39  19  217m  21m 4356 S    0  1.1   0:02.74 beagled-helper
 9607 tom       20   0  190m  16m  11m S    0  0.8   0:01.32 kwrite
 9841 tom       20   0  158m  14m 8620 S    0  0.7   0:00.46 konsole
 4362 tom       27   7  289m  12m 3500 S    0  0.6   2:23.48 beagled
 4335 tom       20   0  198m  10m 6480 S    0  0.5   0:22.82 kicker
 4323 tom       20   0  221m 9824 7000 S    0  0.5   0:18.84 kded
 4333 tom       20   0  194m 9416 6360 S    0  0.5   0:27.18 kdesktop
 8796 tom       20   0  134m 8252 6680 S    0  0.4   0:00.04 konqueror

Load average is 5.39, and apps’ collective memory % looks to be around 55%. The top 5 apps in that list (its sorted by mem%) have VIRT totals nearing 5 gigabytes. CPU usage hovers around 5%ish. Yet, physical memory is reporting only 15mb free? How is that so if %MEM totals around 55%?

I am confused. I just want this box to be responsive like it was before I rebuilt it. I can drop the Win2003 server down to 512mb of RAM. I previously had to disable VMware’s penchant freeing up VM’s memory because it was causing a huge performance hit on the VM’s and the Ubuntu server was missing its backup times.

Help!

No one?

Well, does anyone know why kio_sysinfo needs so much RAM?

No idea. I don’t even have it running, but your’s doesn’t look healthy I’ll agree. A search on the name simply got the info that it is a kio slave to provide system information. Do you have some kind of widget showing system load or something like that. Can you shut it down?

I just noticed that the problem went away last night when I had to kill the vncserver because it locked up. Logging in this afternoon the system acted like nothing was wrong.

Restarting VNC – thereby restarting KDE – and kio_sysinfo doesn’t show up in ps aux. I notice some other kio components do start up, but not sysinfo.

Then I open My Computer under KDE and it presents me with the sysinfo:/ screen. kio_sysinfo appears in the process list but it’s nowhere near the top 10 like it was last night. I’ll leave it on for a bit and see if it does this again.