Open vSwitch Agent ID initalisation prevents reuse of agent with userspace only vSwitches
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
neutron |
Fix Released
|
Low
|
sean mooney |
Bug Description
Problem description
===================
In the Open vSwitch agent, the Agent id is currently based off the mac address of the br-int.
Userspace Open vSwitch derivatives such as the Intel® DPDK Accelerated Open vSwitch do not currently create a tap device in the kernel to back the ovs bridges local port.
This limitation prevents reuse of the OpenVSwitch agent between both switches.
By allowing integration of high throughput vSwitch implementations via existing agents, NFV workloads can be enabled in OpenStack without significant extension of the current codebase.
Proposed change
===============
To enable reuse of the ovs agent with the Intel® DPDK Accelerated Open vSwitch the proposal is to change the generation of the agent_id to use the hostname instead of the mac address of the br-int.
- mac = self.int_
- self.agent_id = '%s%s' % ('ovs', (mac.replace(":", "")))
+ self.agent_id = 'ovs-agent-%s' % socket.
For several plugins such as the nec,mlnx,hyperv and onconvergence agents the hostname is used to create the agent id.
Using the hostname will normalise the agent_id between these 5 neutron agents.
at present the Agent_id is only used in log messages. by using the hostname instead of mac of br-int log readability will also be improved as it will be easier to identify which node the log is from if log aggregation is preformed across a cluster.
Patch
===============
initial patch submitted for review
https:/
Changed in neutron: | |
assignee: | nobody → sean mooney (sean-k-mooney) |
tags: | added: ovs |
Changed in neutron: | |
importance: | Undecided → Wishlist |
status: | New → Confirmed |
Changed in neutron: | |
status: | Confirmed → In Progress |
Changed in neutron: | |
milestone: | none → juno-2 |
importance: | Wishlist → Low |
Changed in neutron: | |
status: | Fix Committed → Fix Released |
Changed in neutron: | |
milestone: | juno-2 → 2014.2 |
Reviewed: https:/ /review. openstack. org/95138 /git.openstack. org/cgit/ openstack/ neutron/ commit/ ?id=744c1bd2f1a e1caf3cd6d0c07d 61b034a21204bc
Committed: https:/
Submitter: Jenkins
Branch: master
commit 744c1bd2f1ae1ca f3cd6d0c07d61b0 34a21204bc
Author: Sean Mooney <email address hidden>
Date: Tue Jun 10 11:41:49 2014 +0100
changes ovs agent_id init to use hostname instead of mac
In the Open vSwitch agent,
the Agent id is currently based off the mac address of the br-int.
Userspace only Open vSwitch derivatives such as Intel's DPDK
Accelerated Open vSwitch do not currently create a tap device in the kernel
to back the ovs bridges local port.
This limitation prevents reuse of the OpenVSwitch agent between both switches.
To enable reuse of the ovs agent with Intel's DPDK Accelerated Open vSwitch,
the proposal is to change the generation of the agent_id to use the hostname,
instead of the mac address of the br-int.
For several plugins such as the nec,mlnx,hyperv and onconvergence agents
the hostname is used to create the agent id.
Using the hostname will normalise the agent_id between these 5 neutron agents,
additionally log readability will also be improved,
if log aggregation is preformed across a cluster
as it will be easier to identify which node the log is from.
the hostname is retrived from cfg.CONF.host
Closes-Bug: #1323259 8f1a17a0931fc98 ac00234ac0b
Change-Id: I9abfac17a74d29