Comment 7 for bug 1532809

Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Related fix merged to nova (master)

Reviewed: https://review.openstack.org/273042
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=32b4ce0e92b333413163f31f2962d109941ee362
Submitter: Jenkins
Branch: master

commit 32b4ce0e92b333413163f31f2962d109941ee362
Author: Sean Dague <email address hidden>
Date: Wed Jan 27 13:02:35 2016 +0000

    Spread allocations of fixed ips

    The default fixed_ip allocation always does .first() out of the fixed
    ip pool. This leads to a pathological situation that in a rapidly
    changing environment we're constantly recycling the same small number
    of ips out of the pool. Given other racey behavior in dnsmasq around
    processing DHCP Release requests, we can start to get DHCP fails when
    a DHCP release is missed, we immediately recycle the IP and give it to
    a new guest.

    We can mitigate this by sorting the results by updated_at. This
    means we'll favor least recently used fixed ips.

    Depends-On: I27f73c1edf12218818c4d279efbd9fef5cdef672

    Related-Bug: #1532809

    Change-Id: I3c83bd68a0e2bbbcdd6d955722dbc9f9fc528113