Jump to: navigation, search

Difference between revisions of "Network/Meetings"

(Agenda for Next Neutron Team Meeting)
(On Demand Agenda)
Line 60: Line 60:
  
 
=== On Demand Agenda ===
 
=== On Demand Agenda ===
 +
* Concrete plan to merge back pecan work
 +
** Need to do this very early after Liberty-2 is cut this week
 
* Vendor decomposition
 
* Vendor decomposition
 
** Which vendors have not started final decomposition?
 
** Which vendors have not started final decomposition?
Line 67: Line 69:
 
*** Currently requirements.txt is left in the vendor dir with the name of the vendor package, but no one seems to know if it is needed.
 
*** Currently requirements.txt is left in the vendor dir with the name of the vendor package, but no one seems to know if it is needed.
 
* Stable branches for networking subproject
 
* Stable branches for networking subproject
** Discussion on this happened in July, decision which wasn't clear was here:
+
** networking sub-projects can perform a release
*** http://lists.openstack.org/pipermail/openstack-dev/2015-July/068785.html
+
** Working with networking-midonet as a test example
** Currently, gerrit ACLs allow only neutron-stable-maint to merge stable patches here
+
** These projects will have a version of 1.0.0
** Stable has meaning, and we have to follow rules of stable backports
 
** Question: What does the team want here?
 
* Neutron interactions with infra
 
** Coordinate through dougwig and mestery
 
** We need to stop adding infra load into #openstack-infra
 
* 3rd party Ci systems: Can they vote -1?
 
** https://review.openstack.org/207198
 
 
* Liberty-3 BPs and reviews
 
* Liberty-3 BPs and reviews
 
** https://launchpad.net/neutron/+milestone/liberty-3
 
** https://launchpad.net/neutron/+milestone/liberty-3
Line 83: Line 78:
 
** We need to merge all of this for Liberty (or make our best shot)
 
** We need to merge all of this for Liberty (or make our best shot)
 
** Email to openstack-dev ML on this topic: http://lists.openstack.org/pipermail/openstack-dev/2015-August/071786.html
 
** Email to openstack-dev ML on this topic: http://lists.openstack.org/pipermail/openstack-dev/2015-August/071786.html
* Concrete plan to merge back pecan and QoS work
 
** Need to do this very early after Liberty-2 is cut this week
 
 
* nova-network compatibility tasks
 
* nova-network compatibility tasks
 
** Summit session(s) in Vancouver were very productive
 
** Summit session(s) in Vancouver were very productive

Revision as of 13:09, 25 August 2015


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

Agenda for Next Neutron Team Meeting

Tuesday (8/25/2015) at 1400 UTC on #openstack-meeting

Announcements / Reminders

Bugs


Important bugs:

    • [1] sporadic test failures due to VMs not getting a DHCP lease
    • [2] neutron test_multi_prefix_slaac failing in the gate with ping failures starting around 7/22
    • [3] neutron.tests.functional.agent.test_l3_agent.MetadataL3AgentTestCase.test_access_to_metadata_proxy times out intermittently
    • [4] Referencing a lb-healthmonitor ID for the first time from Heat would fail
    • [5] migrate_to_ml2 script doesn't work for Juno release
    • [6] security groups don't work with vip and ovs plugin
  • Two bugs are hurting the functional job in the gate:
    • [7] agent.test_l3_agent.MetadataL3AgentTestCase.test_access_to_metadata_proxy times out intermittently
    • [8] radvd exits -1 intermittently in test_ha_router_process_ipv6_subnets_to_existing_port
  • Linuxbridge Gate Parity Bugs
    • [9] nova libvirtError: Unable to add bridge brqxxx-xx port tapxxx-xx: Device or resource busy


Docs (emagana)

Open Items for Liberty: (reviews needed)

Networking Guide

On Demand Agenda

  • Concrete plan to merge back pecan work
    • Need to do this very early after Liberty-2 is cut this week
  • Vendor decomposition
    • Which vendors have not started final decomposition?
      • Should we maintain a list of stragglers in devref until they are all done?
      • Send an email notice about decomposition to -dev: what should it say?
    • Does anything need to be left in the tree for packaging?
      • Currently requirements.txt is left in the vendor dir with the name of the vendor package, but no one seems to know if it is needed.
  • Stable branches for networking subproject
    • networking sub-projects can perform a release
    • Working with networking-midonet as a test example
    • These projects will have a version of 1.0.0
  • Liberty-3 BPs and reviews
  • nova-network compatibility tasks
    • Summit session(s) in Vancouver were very productive
    • Tasks tracked here: https://etherpad.openstack.org/p/YVR-nova-network
    • Need to find owners for a few of these
      • Rolling upgrades (mostly a grenade patch)
      • Distributed SNAT with DVR
      • "Get me a network (https://review.openstack.org/#/c/184857/) (kevinbenton to own this)
      • ARP Spoof patch series is stalled, need to make a decision on how to proceed. (kevinbenton will do this after beating requirements out of markmcclain)
  • macvtap ml2 driver and agent - will land in neutron tree. Question is how to procede with agent code (scheuran)
    • Aug 17 2015 update - sc68cal is working with scheuran in this gerrit topic https://review.openstack.org/#/q/status:open+project:openstack/neutron+branch:master+topic:bug/1468803,n,z
    • 1) Copy (duplicate) required agent code like in the past.
    • 2) Extract a common base class for macvtap, linuxbridge and sriov-nic (impls are close to each other). Purpose: Sharing base agent code. Starting with macvtap only in liberty. Moving others to common base class in Mitaka. New agent base class cannot be used for OVS, as it already diverged too much from others.
      • Liberty: only methods that are obious common and only differ in details that can be abstracted (_setup_rpc, _report_state, _device_info_has_changes, _process_network_devices)
      • Mitaka: Methods that have a lot of common code, but require fruther thinking (treat_devices_added_updated, scan_devices, daemon_loop)
      • N: Having a common agent with interface drivers?
    • Modular l2 agent (spec not in a good shape) and agent extension (is just for extending existing agents) is is not an option

Previous meeting logs