[Openstack] Scalability issue in nova-dhcpbridge

Anton Blanchard anton at samba.org
Mon Mar 19 05:42:42 UTC 2012


Hi,

We are seeing severe boot and delete performance issues with
FlatDHPManager and a lot of instances.

If I have 200 instances running and boot 1 new instance a kill -HUP
dnsmasq calls nova-dhcpbridge 201 times, 200 for the existing leases and
once for the new one.

The 200 events for existing leases end up here in nova-dhcpbridge:

def old_lease(mac, ip_address):
    """Update just as add lease."""
    LOG.debug(_("Adopted old lease or got a change of mac"))
    add_lease(mac, ip_address)

I'm not sure why we need to do this at all. The comment mentions
tracking a change of MAC and yet add_lease doesn't seem to do anything
with the MAC parameter.

I can fix my performance problem by ignoring any old leases, but I was
hoping someone could explain what the purpose is here.

Anton




More information about the Openstack mailing list