Open Stack

*Infrastructure*
Running juno on ubuntu 14.04 with GRE network using  2 network nodes
without DVR.
There are around 200 dhcp namespaces created on each network
nodes(dhcp_network_agents=2) which has 16 cpu and 32 gb.

*Issue:*
Network nodes are idle (top shows 99% idle CPU)but load average is *70-100*
for a network traffic of 100-200 Kbps.
I am not able to figure out why the load average is too high for an idle
server with less than 100 Kbps (including my current ssh session)

there is no impact for the instances but these numbers are worrying.


*some observation:*
CPU time spend on user space is 0%, kernel space is 0.1%,io wait is 0%,swap
usage is 0%,and there are a lot of free RAM,No zombie,No process in un
interruptible sleep.


any idea why load average is very high?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20151114/130c240a/attachment.html>