Back to bug 1394865
| Who | When | What | Removed | Added |
|---|---|---|---|---|
| Itzik Brown | 2016-11-14 15:40:23 UTC | Blocks | 1258832 | |
| Nir Yechiel | 2016-11-14 18:56:15 UTC | Priority | unspecified | high |
| Status | NEW | ON_DEV | ||
| Target Release | --- | 10.0 (Newton) | ||
| Link ID | OpenStack gerrit 357528 | |||
| Assignee | amuller | sgaddam | ||
| Target Milestone | --- | ga | ||
| QA Contact | tfreger | itbrown | ||
| Jon Schlueter | 2016-11-18 19:33:00 UTC | CC | jschluet | |
| Nir Yechiel | 2016-11-30 13:45:13 UTC | Status | ON_DEV | POST |
| Jon Schlueter | 2016-12-01 02:48:34 UTC | Link ID | OpenStack gerrit 357528 | OpenDaylight gerrit 403672 |
| Nir Yechiel | 2016-12-01 13:34:30 UTC | CC | sgaddam | |
| Flags | needinfo?(sgaddam) needinfo?(jschluet) | |||
| Sridhar Gaddam | 2016-12-01 13:40:52 UTC | Link ID | OpenDaylight gerrit 403672 | OpenStack gerrit 403672 |
| Flags | needinfo?(sgaddam) needinfo?(jschluet) | |||
| Jon Schlueter | 2016-12-02 17:17:20 UTC | Keywords | Triaged | |
| Sridhar Gaddam | 2016-12-06 07:57:12 UTC | Doc Text | Cause: OpenStack Neutron supports two modes for ovsdb_interface (i.e., 'vsctl' and 'native') and in stable/mitaka, the default mode used to be 'vsctl'. However, in stable/newton this was changed to 'native'[2]' interface. In 'native' mode, when using OpenStack (i.e., Newton release) with external controllers (like ODL), it was noticed that dhcp-agent was overriding the OVS manager on the switch to loopback ip (i.e., default value of ovsdb_connection). Consequence: Since the dhcp-agent overrides the manager on the switch, the connection between the existing ODL controller and the switch is lost and creates all kinds of issues. Fix: Fix Neutron not to override the manager on the switch, but rather update it by adding a new manager as necessary. Result: With the fix in Neutron, dhcp-agent simply adds a new manager to the switch rather than overriding the existing manager. This way, the connection between the switch and external controller is not disturbed. | |
| Doc Type | If docs needed, set a value | Bug Fix | ||
| Ihar Hrachyshka | 2016-12-13 14:07:46 UTC | CC | ihrachys | |
| RHEL Program Management | 2016-12-13 14:16:40 UTC | Keywords | ZStream | |
| Scott Lewis | 2016-12-20 16:10:54 UTC | Target Milestone | ga | --- |
| Sridhar Gaddam | 2016-12-20 16:14:03 UTC | Status | POST | ON_QA |
| Jon Schlueter | 2016-12-21 03:31:20 UTC | Status | ON_QA | POST |
| Jon Schlueter | 2017-01-06 23:06:09 UTC | Status | POST | MODIFIED |
| Fixed In Version | openstack-neutron-9.1.1-2.el7ost | |||
| Nir Yechiel | 2017-01-08 06:30:08 UTC | Flags | needinfo?(jschluet) | |
| Nir Yechiel | 2017-01-08 07:05:05 UTC | Target Milestone | --- | z1 |
| errata-xmlrpc | 2017-01-13 14:19:55 UTC | Status | MODIFIED | ON_QA |
| Don Domingo | 2017-01-17 12:48:04 UTC | CC | ddomingo | |
| Doc Text | Cause: OpenStack Neutron supports two modes for ovsdb_interface (i.e., 'vsctl' and 'native') and in stable/mitaka, the default mode used to be 'vsctl'. However, in stable/newton this was changed to 'native'[2]' interface. In 'native' mode, when using OpenStack (i.e., Newton release) with external controllers (like ODL), it was noticed that dhcp-agent was overriding the OVS manager on the switch to loopback ip (i.e., default value of ovsdb_connection). Consequence: Since the dhcp-agent overrides the manager on the switch, the connection between the existing ODL controller and the switch is lost and creates all kinds of issues. Fix: Fix Neutron not to override the manager on the switch, but rather update it by adding a new manager as necessary. Result: With the fix in Neutron, dhcp-agent simply adds a new manager to the switch rather than overriding the existing manager. This way, the connection between the switch and external controller is not disturbed. | The OpenStack Networking service supports two modes for ovsdb_interface, namely: 'vsctl' and 'native'. In Red Hat OpenStack Platform 9, the default mode was 'vsctl'. In the last version, the default changed to 'native'. In 'native' mode, when using Red Hat OpenStack Platform 10 with external controllers (such as ODL), the dhcp-agent overrode the OVS manager on the switch to loopback IP (ie. the default value of ovsdb_connection). As a result, the connection between the existing controller and the switch was lost. With this release, OpenStack Networking no longer overrides the manager on the switch. Rather, a new manager is added as necessary. This way, the connection between the switch and external controller stays intact. |
||
| Itzik Brown | 2017-01-18 13:58:21 UTC | Status | ON_QA | VERIFIED |
| Flags | needinfo?(jschluet) | |||
| errata-xmlrpc | 2017-01-25 11:18:41 UTC | Status | VERIFIED | RELEASE_PENDING |
| errata-xmlrpc | 2017-01-26 16:41:40 UTC | Status | RELEASE_PENDING | CLOSED |
| Resolution | --- | ERRATA | ||
| Last Closed | 2017-01-26 11:41:40 UTC | |||
| Mike Kolesnik | 2018-10-18 07:19:13 UTC | Environment | N/A |
Back to bug 1394865