We are seeing strange behaviour on an Ubuntu 8.04 (Hardy x64) guest with vmware tools installed.
The kernel version is 2.6.24-19-server x86_64
On boot, the guest has a high number of context switches, tens of thousands:
procs -
memory----
-swap -
io---- -system-- -
cpu----
r b swpd free inact active si so bi bo in cs us sy id wa
2 0 0 1996924 19664 18868 0 0 1503 41 105 26891 9 21 70 1
0 0 0 1996924 19656 18884 0 0 0 0 15 65524 0 0 100 0
0 0 0 1996924 19656 18884 0 0 0 0 9 70856 0 0 100 0
0 0 0 1996924 19656 18892 0 0 0 0 7 86828 0 0 100 0
0 0 0 1996924 19656 18892 0 0 0 0 9 77066 0 0 100 0
0 0 0 1996924 19656 18892 0 0 0 0 7 83680 0 0 100 0
0 0 0 1996924 19656 18892 0 0 0 16 13 82136 0 0 100 0
After restarting vmware-guestd, the problem disappears, and the guest behaves as expected:
procs -
memory----
-swap -
io---- -system-- -
cpu----
r b swpd free inact active si so bi bo in cs us sy id wa
1 0 0 1362668 41168 590980 0 0 50 6 13 36 0 0 99 0
0 0 0 1362396 41168 590968 0 0 0 0 20 63 0 1 99 0
0 0 0 1362396 41168 590968 0 0 0 56 10 37 0 0 100 0
0 0 0 1362396 41168 590968 0 0 0 0 20 27 0 0 100 0
0 0 0 1362404 41168 590968 0 0 0 0 5 27 0 0 100 0
0 0 0 1362404 41168 590968 0 0 0 0 19 29 0 0 100 0
Additionally, the problem does not occur if vmware-tools is not installed. The problem does not occur in older Ubuntu distributions (ie Feisty, Gutsy)
Although the number of context switches is high, the server does not seem to be using up vCPU resources while idling.
Has anybody else observed this behaviour?
