VLAN not tagged on every switchs

[Gaëtan Trellu]

When a Neutron network is created, the VLAN is propagated on every switch but when we are provisioning a baremetal node via Nova/Ironic the VLAN is only tagged on one switch only.

We have a baremetal node using the port group implemented in Ocata.

SW11(config-if-Po116)#sh active
interface Port-Channel116
   switchport access vlan 1100
   switchport trunk allowed vlan none
   port-channel lacp fallback static
   port-channel lacp fallback timeout 5
   mlag 116

SW12(config-if-Po116)#sh active
interface Port-Channel116
   switchport trunk allowed vlan none
   port-channel lacp fallback static
   port-channel lacp fallback timeout 5
   mlag 116

On SW12 switch the VLAN 1100 is not tagged for the Port-Channel116 interface. With this behavior our LACP setup doesn't work.

We are running Pike with NGS 1.0.0 on Arista switches.

 $openstack baremetal port show 8b5bdf59-9fb5-4afe-b467-89370b51bfbc -c local_link_connection -f json | jq .
{
  "local_link_connection": {
    "switch_info": "SW11",
    "port_id": "Port-Channel116",
    "switch_id": "00:1c:73:e2:15:27"
  }
}

$ openstack baremetal port show a977a1c3-18f7-46ac-bf20-eeb39c4c5613 -c local_link_connection -f json | jq .
{
  "local_link_connection": {
    "switch_info": "SW12",
    "port_id": "Port-Channel116",
    "switch_id": "00:1c:73:e2:15:ed"
  }
}

$ openstack baremetal port group show 3fb3c2db-6058-4ddc-a6d5-b723802f4009 -c internal_info -f json | jq .
{
  "internal_info": {
    "tenant_vif_port_id": "0562ecb1-7c90-4585-9d48-30c08bc15932"
  }
}

$ openstack port show 0562ecb1-7c90-4585-9d48-30c08bc15932 -c binding_profile -f json | jq .
{
  "binding_profile": "local_group_information='{u'bond_mode': u'802.3ad', u'id': u'3fb3c2db-6058-4ddc-a6d5-b723802f4009', u'name': u'CZ35359C3LBond', u'bond_properties': {u'bond_lacp_rate': u'fast', u'bond_updelay': u'200', u'bond_xmit_hash_policy': u'layer3+4', u'bond_miimon': u'100', u'bond_downdelay': u'200'}}', local_link_information='[{u'switch_info': u'SW11', u'port_id': u'Port-Channel116', u'switch_id': u'00:1c:73:e2:15:27'}, {u'switch_info': u'SW12', u'port_id': u'Port-Channel116', u'switch_id': u'00:1c:73:e2:15:ed'}]'"
}

Looks like there is no loop in local_link_information and it took the first switch it found, here SW11 which is the one with the VLAN correctly tagged.

[Gaëtan Trellu (goldyfruit)]

I added few information about the Ironic and Neutron ports.

[Gaëtan Trellu (goldyfruit)]

In the logs I can see the provisioning network VLAN tagged on both switches but for the tenant network VLAN it only happened on SW11.

$ grep -ri "Successfully bound port"
neutron-server.log.1:2018-03-26 18:19:55.811 27 INFO networking_generic_switch.generic_switch_mech [req-1596620f-b5bb-4925-8876-994458e31c43 a1202baec4954680ab2ac3bc4b9c39c2 0b096e2e85b64812b7ed348d8a6f3fd5 - default default] Successfully bound port 6b44c197-0327-449b-9d79-604f6d72ad1f in segment 65 on device SW11

neutron-server.log.1:2018-03-26 18:20:09.384 27 INFO networking_generic_switch.generic_switch_mech [req-8dfd8a7a-5150-4878-b59d-29aea1ab55e7 a1202baec4954680ab2ac3bc4b9c39c2 0b096e2e85b64812b7ed348d8a6f3fd5 - default default] Successfully bound port 908821df-ef7e-4557-b4e7-532c5a329934 in segment 65 on device SW12

neutron-server.log.1:2018-03-26 18:28:44.912 26 INFO networking_generic_switch.generic_switch_mech [req-ce807910-c956-4ac6-ad93-faa6db6ae753 a1202baec4954680ab2ac3bc4b9c39c2 0b096e2e85b64812b7ed348d8a6f3fd5 - default default] Successfully bound port 0562ecb1-7c90-4585-9d48-30c08bc15932 in segment 1100 on device SW11

[Gaëtan Trellu (goldyfruit)]

I guess this is the same issue: https://bugs.launchpad.net/networking-generic-switch/+bug/1576597

[Gaëtan Trellu (goldyfruit)]

https://review.openstack.org/#/c/557051/

[Gaëtan Trellu (goldyfruit)]

After applying the change above I can see the VLAN tagged on bot switches.

2018-03-27 20:35:19.236 44 INFO networking_generic_switch.generic_switch_mech [req-e29ebf32-f89c-4d7e-9f83-b3ca9e159753 a1202baec4954680ab2ac3bc4b9c39c2 0b096e2e85b64812b7ed348d8a6f3fd5 - default default] Successfully bound port 3588a1ee-18e5-479a-9074-8103a72487c3 in segment 65 on device SW11

2018-03-27 20:35:32.624 44 INFO networking_generic_switch.generic_switch_mech [req-6ea17d8d-9d22-4ffd-9e19-f64fb2572e2a a1202baec4954680ab2ac3bc4b9c39c2 0b096e2e85b64812b7ed348d8a6f3fd5 - default default] Successfully bound port b8a4b250-ea9b-4ea8-9e02-3d3bfc7429d3 in segment 65 on device SW12

2018-03-27 20:43:44.218 44 INFO networking_generic_switch.generic_switch_mech [req-908c61af-54c7-45cd-9ac8-031d93656cfb a1202baec4954680ab2ac3bc4b9c39c2 0b096e2e85b64812b7ed348d8a6f3fd5 - default default] Successfully bound port 2168b48e-6b7a-4993-8e52-6238250be7fa in segment 1100 on device SW11
2018-03-27 20:43:55.638 44 INFO networking_generic_switch.generic_switch_mech [req-908c61af-54c7-45cd-9ac8-031d93656cfb a1202baec4954680ab2ac3bc4b9c39c2 0b096e2e85b64812b7ed348d8a6f3fd5 - default default] Successfully bound port 2168b48e-6b7a-4993-8e52-6238250be7fa in segment 1100 on device SW12

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/networking-generic-switch 6.1.0 release.