RN6.1 Neutron features
This patch adds information about Neutron features in the MOS 6.1. Change-Id: If6ac4736d7d1e066157783be8f249137857bc54a
This commit is contained in:
committed by
Maria Zlatkova
parent
f527493a4b
commit
c83f97ee96
@@ -1,4 +1,3 @@
|
||||
|
||||
.. _fuel-network.rst:
|
||||
|
||||
Networking Issues
|
||||
@@ -28,4 +27,3 @@ Known networking issues
|
||||
See `LP1457478 <https://bugs.launchpad.net/fuel/+bug/1457478>`_.
|
||||
|
||||
.. include:: /pages/release-notes/v6-1/9100-mellanox.rst
|
||||
|
||||
|
||||
@@ -41,7 +41,6 @@ Issues in OpenStack Components
|
||||
of a Kubernetes cluster.
|
||||
See `LP1461564 <https://bugs.launchpad.net/fuel/+bug/1461564>`_.
|
||||
|
||||
|
||||
.. include:: /pages/release-notes/v6-1/other/2010-general.rst
|
||||
.. include:: /pages/release-notes/v6-1/other/3131-neutron.rst
|
||||
.. include:: /pages/release-notes/v6-1/other/7010-nova.rst
|
||||
|
||||
@@ -1,3 +1,16 @@
|
||||
Multiple L3 Agents per environment
|
||||
----------------------------------
|
||||
|
||||
Fuel 6.1 deploys one :ref:`L3 Agent<l3-agent-term>` per
|
||||
Controller node, which helps eliminate HA bottlenecks
|
||||
that could occur in environments with only one L3 agent running.
|
||||
Rescheduling of networks is moved to Neutron server code.
|
||||
Neutron server code reschedules networks
|
||||
by automatically reassigning routers to L3 agents
|
||||
when it detects that a particular L3 agent is failed.
|
||||
See :ref:`tshoot-corosync-ops` for examples and the
|
||||
`fuel-multiple-l3-agents blueprint <https://blueprints.launchpad.net/fuel/+spec/fuel-multiple-l3-agents>`_
|
||||
for details about the implementation.
|
||||
|
||||
Neutron agent state reporting
|
||||
-----------------------------
|
||||
@@ -8,9 +21,31 @@ REST API calls to notify a neutron server
|
||||
which maintains agents' state by collecting state
|
||||
reports from agents via AMQP. They can report their
|
||||
own status by saving it in local files.
|
||||
So, when a message queue has issues the Cluster Resource Manager
|
||||
So, when a message queue has issues, the Cluster Resource Manager
|
||||
still can respond in time if something goes
|
||||
wrong with an agent.
|
||||
|
||||
See the `Neutron agents blueprint <https://blueprints.launchpad.net/fuel/+spec/neutron-agents-local-reports>`_
|
||||
wrong with an agent. See the `neutron-agents-local-reports blueprint
|
||||
<https://blueprints.launchpad.net/fuel/+spec/neutron-agents-local-reports>`_
|
||||
for details about the implementation.
|
||||
|
||||
.. note::
|
||||
This feature is disabled by default in Mirantis OpenStack 6.1
|
||||
since it solves a very specific corner case, which might happen
|
||||
in production. So enabling it by default is very risky and
|
||||
should be handled with the help of Mirantis support team.
|
||||
|
||||
Multiple DHCP-agents
|
||||
--------------------
|
||||
|
||||
In Mirantis OpenStack 6.1, multiple DHCP agents are configured with
|
||||
one on each Controller. In earlier releases, each environment
|
||||
had a single DHCP agent that was located on one of the Controllers.
|
||||
With the help of multiple DHCP agents, you can avoid the HA
|
||||
bottlenecks that could occur with only one DHCP agent running in the
|
||||
environment. Also, each network with a DHCP server enabled is served
|
||||
by two DHCP agents simultaneously, so a failure of one DHCP agent is
|
||||
completely transparent to DHCP clients. Rescheduling of networks is
|
||||
moved to Neutron server code, which accomplishes this by
|
||||
automatically reassigning networks to DHCP agents when it detects
|
||||
that a particular DHCP agent is dead. See the `fuel-multiple-dhcp-agents blueprint
|
||||
<https://blueprints.launchpad.net/fuel/+spec/fuel-multiple-dhcp-agents>`_
|
||||
for details about the implementation.
|
||||
|
||||
@@ -3,7 +3,32 @@
|
||||
OpenStack networking (Neutron)
|
||||
------------------------------
|
||||
|
||||
Resolved Neutron issues
|
||||
Neutron features supported in 6.1
|
||||
+++++++++++++++++++++++++++++++++
|
||||
|
||||
* DB migration refactor and new timeline.
|
||||
|
||||
* IPSet support for security groups (this option is configurable).
|
||||
|
||||
* L3 agent performance improvements such as processing many routers
|
||||
in parallel and improved responsiveness to L3 changes made through
|
||||
the API following an agent restart.
|
||||
|
||||
* Migration to oslo.messaging library for RPC communication.
|
||||
|
||||
* Security group rules for devices RPC call refactoring (a huge
|
||||
performance improvement).
|
||||
|
||||
Neutron features not supported in 6.1
|
||||
+++++++++++++++++++++++++++++++++++++
|
||||
|
||||
* Distributed Virtual Router Support (DVR).
|
||||
|
||||
* Full IPv6 support for tenant networks.
|
||||
|
||||
* High Availability for the L3 Agent.
|
||||
|
||||
Resolved Neutron Issues
|
||||
+++++++++++++++++++++++
|
||||
|
||||
* In rare circumstances, there was no connectivity to instances in HA
|
||||
|
||||
Reference in New Issue
Block a user