Fix haproxy agent unit test to be runnable alone by tox
haproxy agent uses periodic_interval option that can interfere
with the same option declared in neutron/service.py when running
'tox services.loadbalancer'
Bob Kukura [Fri, 13 Sep 2013 20:12:30 +0000 (16:12 -0400)]
Use same hostname function as nova
Certain plugin implementations of port binding depend on the port's
binding:host_id set by nova matching the compute node's hostname
reported by neutron agents in agents_db. Nova uses
socket.gethostname() to obtain the hostname, so this changes neutron
to also use socket.gethostname(), instead of socket.getfqdn() to
obtain the default hostname. If deployments require a specific
hostname format or value, the hostname can be specified in nova and
neutron configuration files.
Prevent stale DHCP directories for dhcp agent nodes
This change fixes an issue that is most acute in test/dev
environments but that may affect production environments
as well: the presence of DHCP directories that no longer
represent networks that exist in the Neutron DB. This
issue can manifest itself over time if you delete networks
from the Server while the Agent node is down.
Without this fix, at the agent start-up the method
existing_dhcp_networks will return an empty list, and
as a consequence the directories that belong to networks
that have been removed from the server will stay because
the sync logic will not process them as it only looks at
the mismatch between active networks on the agent and
active networks on the server. With the fix instead, we
return *all* existing dhcp networks; if they are meant to
be active, the agent will bring them up, if they no longer
exist on the server, then the agent will dispose of the
resources. The dnsmasq driver will do the right thing when
enabling or disabling the process.
pass default l3_gateway_service_uuid if not specified
Previously, if one created an external network of type l3_ext and did
not specify the default_gw_service_uuid via --provider:physical_network
an exception would be raised on the backend as we need to specify the
l3_gw_service_uuid we want nvp to use.
Luis A. Garcia [Fri, 16 Aug 2013 16:07:47 +0000 (16:07 +0000)]
Enhance exception translation to better handle NeutronExceptions
NeutronExceptions have a 'message' class attribute that holds the
generic error message template, e.g. "Network %(network)s not found",
unfortunately, because the names are the same, it was overshadowing the
actual exception instance 'message', e.g. "Network 1 not found", after
translation when the exception was serialized to JSON.
This patch puts the exception's actual message in a new field called
'msg' and overwrites NeutronException unicode() so that 'msg' is used
during serialization and we'll get the correct message on the REST API
response.
Add a relationship with eager loading in the Port and Network models,
thus preventing the 'extend' functions from performing extra queries.
This patch also slight alters the methods for processing qos_queue
bindings in order to allow them for populating the qos_queue id in
the response being created.
This patch introduces VXLAN support for Linuxbridge agent alongside
with ml2 plugin support in linubridge mechnism driver.
A new vxlan configuration section is added for vxlan related parameters.
The agent also implements l2population RPC callbacks which allows ml2
plugin using l2population mechnism driver to populate vxlan forwarding
and neighbor tables following portbinding events. It allows agent to
respond locally to ARP requests for remote VMs and avoid dataplane based
learning. This should help limiting the use of multicast or flooding
for broadcast emulation in vxlan networks.
These changes should anyway have a limited risk, as agent behaviour
shouldn't be affected, except when vxlan is enabled alongside ml2 plugin.
Francois Eleouet [Thu, 22 Aug 2013 14:51:01 +0000 (16:51 +0200)]
OVS agent implementation of l2-population
This patchset implements l2-population RPC callbacks in OVS agents,
it enables plugin to populate forwarding table following portbindings
events.
For now, it doesn't include ARP responder implementation which is
deferred to a future patchset (As this feature isn't yet supported by
OVS, it will require the use of an external responder such as ebtables)
It anyway brings some improvements in tunnelling management, as agent
will tear-down unecessary tunnels, and flood packets on a per-network
basis rather than to all other agents.
These changes should anyway have a limited risk, as tunnel management
won't be affected as long as l2_population option is not set. This
option must be used in conjonction with ml2 plugin using l2population
mechanism driver.
Kui Shi [Wed, 11 Sep 2013 17:04:09 +0000 (01:04 +0800)]
Don't need to init testr in run_tests.sh
In run_tests.sh, function init_testr will initialize testr if the
directory .testrepository is not existed. Actually, testr will do
the check before run the test:
In Python package testrepository, setuptools_command.py:Testr.run
68 def run(self):
69 """Set up testr repo, then run testr"""
70 if not os.path.isdir(".testrepository"):
71 self._run_testr("init")
Bob Melander [Wed, 3 Apr 2013 19:22:30 +0000 (21:22 +0200)]
Adds support for L3 routing/NAT as a service plugin
- Adds L3 routing/NAT service plugin
- Removes L3 routing/NAT from ML2 plugin
- Moves "router:external" attribute to new extension "External-net"
- Introduces separate RPC topic for L3 callbacks from L3 agent
Kaiwei Fan [Fri, 23 Aug 2013 06:25:52 +0000 (23:25 -0700)]
Support for NVP advanced service router
When creating an LR:
- deploy an Edge asynchronously
- create a L2 switch for connecting LR and Edge
- attach a router port to the L2 switch.
- assign ip address 169.254.2.1/28 and nexthop 169.254.2.3 to LR
When set external gateway:
- configure Edge interface and default gateway
- Add static routes to Edge for all logic networks attached to LR via nexthop 169.254.2.1
- configure SNAT rules for all logic networks attached to LR
When add router interface:
- Add static route/SNAT rule for the network attached to LR
When associate floating IP address:
- configure DNAT rule for the floating ip and the port
Tests being done:
- Verified Edge is deployed asynchronously and LR is attached to the internal created L2 switch
- Manually attach Edge's vNic to the L2 switch and Edge is able to ping 169.254.2.1
- Verified router-delete deletes Edge asynchronously and remove the internal L2 switch
- Verified SNAT/DNAT/static-routes rules are configured on Edge in correct order
- Verified external vnic ip address/netmask and default gateway is configured
Help text was added to the configuration options defined in the brocade neutron
plugin. This help text should assist users of the brocade plugin. The OSTYPE
configuration option was not used in the plugin and was labeled as such so a
user of the plugin would not worry about the meaning of it.
The example etc file /etc/neutron/plugins/brocade/brocade.ini has been updated
to share the same help values where applicable.
Ensure pid file is removed when metadata ns daemon receives SIGTERM
These files from the metadata namespace proxy are not being removed
because delete_pid() is registered with atexit. This means it only runs
when a process exits normally and won't run when a process receives a
signal.
This patch registers a signal handler for SIGTERM that calls exit()
to make the process exit normally so delete_pid() gets called.
LBaaS: Fix healthmonitor disassociation for non-admin
Due to specifics of policy engine, checked object should have
tenant_id to be checked by rule admin_or_owner.
In 'disassociate' operation neutron API layer works with
PoolHealthMonitorAssociation which doesn't have tenant_id field.
Need to add it to resulting dict returned by get_pool_health_monitor.
Using tools/check_i18n.py to scan source directory, and fix most of
the errors.
- Message internationalization
- First letter must be capital
- Using comma instead of percent in LOG.xxx
Note: all extension's description are not touched in this patch,
can be fixed after discussing.
Note: all nicira/check_nvp_config.py print messages are not fixed.
Kaiwei Fan [Thu, 5 Sep 2013 20:57:13 +0000 (13:57 -0700)]
Fix IF checks on spawned green thread instance
Initially the symptom looks like race condition between two threads when
stopping the task manager. After further analysis/troubleshooting, it
turns out that two threads are spawned if a task manager is stopped and
started again, causing unexpected errors.
The IF check on the spawned thread sometimes return True sometime return False
if not compared against None explicitly. This makes start() method
think no thread has been started or stop() method think no thread is started.
Change the check to compare against None.
Also fixed a problem in unit-test where a thread may never terminated when
a stop call is invoked during db access.
Prevents 400 NVP errors caused by a None display_name
The API forbids a resource name to be None, but the
Model does not. Such errors may be induced by
programming directly against the plugin interface. With
this fix we avoid raising 400 faults which may be introduced
by involuntary programming errors.
neutron.common.log.log is useful for logging arguments of a method.
It outputs class name and method name, but module path is not output.
A module path is useful to search the log message.
stevedore requires an additional parameter to be used (name_order=True) to
sort the loaded extensions to match the order used in the parameter "names".
ZhiQiang Fan [Fri, 6 Sep 2013 05:27:19 +0000 (13:27 +0800)]
Fix incorrect NotImplementedError
neutron.common.exceptions doesn't define NotImplementedError, but
some codes still use it. We should use builtin
exceptions.NotImplementedError instead.
Note: this patch also fixes a never run code in nvp.
Code Review / openstack-build / neutron-build.git / log