Jump to: navigation, search


Revision as of 21:45, 19 December 2014 by Anteaya (talk | contribs) (On Demand Agenda)

Meeting time: The OpenStack Networking Team (Neutron) holds public meetings alternating between Mondays at 2100 UTC (#openstack-meeting) and Tuesdays at 1400 UTC (#openstack-meeting). Everyone is encouraged to attend.

Apologies for Absence

  • Brent Eagles (beagles) - I've volunteered tech skills for an event. I've added some items to the bugs section.

Agenda for Next Neutron Team Meeting

Tuesday (12/16/2014) at 1400 UTC on #openstack-meeting

Announcements / Reminders


New issues discovered recently/having most hits in the gate:

    • [1] Race to delete shared subnet in Tempest neutron full jobs
    • [2] Failure to allocate tunnel id when creating networks concurrently
    • [3] br-tun lost ports/flows if openvswitch restart
    • [4] Security group rules errorneously applied to all ports having same ip addresses in different networks
    • [5] ping still working after security group rule is created, updated, or deleted (related to previous one)

These bugs are in nova but are related to neutron. It would be great if we could get neutron reviews on:

    • Merged [6] Remove unneeded call to fetch network info on shutdown - arosen
    • Merged [7] remove unneeded call to network_api on detach_interface - arosen
    • Merged [8] remove unneeded call to network_api on rebuild_instance - arosen
    • Merged [9] Optimize validate_networks to query neutron only when needed - arosen
    • [10] deallocate_for_instance should delete all neutron ports on error - arosen
    • [11] Fix port_security_enabled neutron extension - arosen
    • [12] Fix pre-created ports in neutron from being deleted by nova - arosen

This issue is still happening

As suggested in last weeks meeting, there is a new BZ https://bugs.launchpad.net/neutron/+bug/1398566. This needs confirmation so it can be given appropriate status. In the live systems where it shows up it is rather awkward to work around as you basically have to shutdown all of the agents before restarting the API server. I've put a patch up for review.

The steps to reproduce are very timing and order dependent. I've created video demonstration on how I've simulated a loaded environment and the steps to reproduce, a discussion of Vincent Untz's patch that my patch is derived from and a demonstration of the fix http://youtu.be/-nL1gyL3KL8 and http://youtu.be/XUY82_a7k5Q . I apologize for the blurry text quality - but the HD transcoding is completed now so if you set the playback to 1080p, it is what you see IRL.

Docs (emagana)

Open Items for Kilo:

Networking Guide:

On Demand Agenda

  • Services split update
    • http://lists.openstack.org/pipermail/openstack-dev/2014-December/052713.html
    • Devstack working
    • 3rd party CIs using devstack-gate should now be working, others need to be updated
    • Please re-submit reviews (lbaasv2 reviews, ask in #openstack-lbaas for status and which dependent review)
    • Moving extensions, config files/items, and separating tempest tests remain and will be part of Kilo
    • Please file bugs for any issues
  • Nova-Network to Neutron Migration
    • anteaya to give an update

Previous meeting logs