Hello,
I am having a problem with my icehouse testbed, which has one
controller, one network node (OVS/VLAN) and ~50 compute nodes.
It runs fine for some time, but then recently, only a small portion of
new instances can start, most of the new instances are stuck in BUILD
forever, without an IP assigned. The scheduler log shows the scheduler
has chosen the compute node to start the instance, but the neutron
server log shows no sign of allocating MAC/IP addresses for it, I
suspect the neutron server didn't get the request for some reason.
Any suggestions on how to debug this?
Thanks,
Xin