Kevin Benton [Thu, 12 Mar 2015 01:32:52 +0000 (18:32 -0700)]
Schedule net to a DHCP agt on subnet create
Change the DHCP notifier behavior to schedule a network
to a DHCP agent when a subnet is created rather than
waiting for the first port to be created.
This will reduce the possibility to get a VM port created
and have it send a DHCP request before the DHCP agent is
ready. Before, the network would be scheduled to an agent
as a result of the API call to create the VM port, so the
DHCP port wouldn't be created until after the VM port.
After this patch, the network will have been scheduled to
a DHCP agent before the first VM port is created.
There is still a possibility that the DHCP agent could be
responding so slowly that it doesn't create its port and
activate the dnsmasq instance before the VM sends traffic.
A proper fix will ensure that the dnsmasq instance is
truly ready to serve requests for a new port will require
significantly more code for barriers (either on the subnet
creation, port creation, or the nova boot process) are too
complex to add this late in the cycle.
This patch also eliminates the logic in the n1kv plugin that
was already doing the same thing.
This patch introduces an issue with the Tempest test. Already there was a similar patch that broke the tempest test for DVR.
This patch consistently failed to pass test_add_list_remove_router_on_l3_agent.
I would recommend to revert this patch until the tempest test is fixed.
Assaf Muller [Thu, 12 Mar 2015 23:50:24 +0000 (19:50 -0400)]
Move Unix domain socket helpers to a common place
As part of the all consuming report-ha-router-master, a new
per router neutron-keepalived-state-change daemon will alert
the L3 agent on every keepalived state change. Since it will
use the Unix domain socket helpers, and they're currently
located in metadata related places, this patch moves them
to a common location.
Also, the UnixDomainHTTPConnection connection string
may now be overridden.
ChuckC [Mon, 9 Mar 2015 00:51:09 +0000 (17:51 -0700)]
Prevent updating mac address of bound port
Currently, a port's mac address can be updated even if it is bound.
This fixes the _check_mac_update_allowed() call to pass just port update
attributes rather than the entire payload and fills in missing testing.
Maru Newby [Fri, 13 Mar 2015 18:17:29 +0000 (18:17 +0000)]
Set TEMPEST_CONFIG_DIR in the api tox env
TEMPEST_CONFIG_DIR needs to be set for the api job to execute
successfully. Previously this was being set by the gate but local
execution required manually setting the variable.
Sergey Belous [Mon, 9 Feb 2015 16:38:05 +0000 (19:38 +0300)]
Reduce db calls count in get_devices_details_list
Each Neutron agent will impose db calls to Neutron Server
to query devices, port and networks.
Network caching is added to reduce the number
of db calls on get_devices_details_list.
Added unit tests for the check caching.
Shiv Haris [Tue, 3 Feb 2015 01:58:53 +0000 (17:58 -0800)]
Brocade vendor code decomposition from neutron repo.
Brocade code decomposition of VDX mechanism driver,
includes moving out the driver and unit tests. Left
behind are the DB model for migration and config
Tim Swanson [Fri, 6 Feb 2015 23:49:49 +0000 (18:49 -0500)]
Add MTU selection & advertisement settings to Neutron config
neutron.conf parameter additions in support of MTU selection
and advertisement. Also, under ML2 group, add an additional
configuration params to optionally set the path and segment
MTU values as well as optionally map physnets to MTU values.
lzklibj [Thu, 26 Feb 2015 05:19:07 +0000 (21:19 -0800)]
fix check_ports_exist_on_l3agent in no subnet case
If no subnets attached to the given router, this check
should return False.
Currently, if no subnets attached to given router, the
following process in this method will fetch all ports
to continue its checking, consider those ports are not
related to the given router, the following checking
should be invalid.
To issue #1378066, after running "router-gateway-clear",
_schedule_router will be triggered, and the invalid
checking will make processing in get_candidates believe
that all l3-agents are valid to schedule this router,
and finally, invalid records are inserted into table
RouterL3AgentBindings.
Russell Bryant [Tue, 9 Dec 2014 21:59:20 +0000 (14:59 -0700)]
Set default of api_workers to number of CPUs
The current default for api_workers is 0. This patch changes it to
use the number of CPUs available, which is the default that will
provide best performance. This is consistent with some other
OpenStack service (Nova and Swift, at least, perhaps others).
Note that devstack is updated in another patch to ensure multiple API
workers is being tested for Neutron the same way it's tested for
several other services:
Eugene Nikanorov [Thu, 22 Jan 2015 12:54:29 +0000 (15:54 +0300)]
Refactor retry mechanism used in some DB operations
Use oslo_db helper that will allow to restart the whole
transaction in case it needs a certain operation to be repeated.
This is a workaround for the REPEATABLE READ problem where
retrying logic will not work because queries inside a transation
will not see updates made by other transactions.
So, run every attempt in a separate transaction.
Terry Wilson [Mon, 23 Feb 2015 20:56:44 +0000 (14:56 -0600)]
Add rootwrap daemon mode support
This patch introduces support for rootwrap daemon mode. It adds
a new config option, AGENT.root_helper_daemon with no default. To
enable, set to something like:
The patch currently assumes that the root_helper_daemon value, and specifically
the rootwrap config, will not change once calls to execute() happen. While it
would not be hard to generate a rootwrap daemon client for each new config, I
couldn't think of a legitimate reason to support it and left it out as YAGNI.
This patch does change the behavior of the addl_env argument to create_process
and execute. Previously, an environment dict would be passed to Popen. If
a root helper was used, this environemnt would actually be passed to 'sudo'
which would filter it before passing it to the underlying command. In the case
of daemon mode, this would cause a problem as the enviornment is filtered by
sudo only once, at daemon startup. Any environment variables added at execute
time would then just be passed directly to the underyling command unfiltered.
oslo.rootwrap 1.6.0 fixes this issue by denying the passing of environment
variables to the daemon altogether. Instead, anything using rootwrap and needing
to pass additional environment variables should define an EnvFilter and run the
command with env var=val cmd. utils.execute/create_process have been modified to
run code in this way (which netns.execute already did).
No code in neutron currently uses both run_as_root=True and addl_env, so this
change does not require any change in code or filters.
Break coupling between ML2 and L3 during create/update operations
This is an initial patch in a series that, by using an event framework,
cleans up the relationship between ML2 and L3, so that they are no longer
tightly coupled. A follow-up will take address the coupling during the
port delete operation.
The newly introduced notification hooks not only benefit the L3 service
plugin, but any other plugin that has an interest in knowing about port
events.
Long term, the notification bits can move in a more 'common' place so that
other plugins can take advantage of them, but as mentioned in a parent patch,
the perestroika is not quite there yet.
Ihar Hrachyshka [Thu, 26 Feb 2015 22:26:41 +0000 (23:26 +0100)]
Migrate to oslo.log
It's mostly a matter of changing imports to a new location.
Non-obvious changes needed:
* pass overwrite= argument to oslo_context since oslo.log reads context
from its thread local store and not local.store from incubator
* don't store context at local.store now that there is no code that
would consume it
* LOG.deprecated() -> versionutils.report_deprecated_feature()
* dropped LOG.audit check from hacking rule since now the method does
not exist
* WritableLogger is now located in oslo_log.loggers
Dropped log module from the tree. Also dropped local module that is now
of no use (and obsolete, as per oslo team).
Added versionutils back to openstack-common.conf since now we use the
module directly from neutron code and not just as a dependency of some
other oslo-incubator module.
Note: tempest tests are expected to be broken now, so instead of fixing
all the oslo.log related issues for the subtree in this patch, I only
added TODOs with directions for later fix.
Assaf Muller [Thu, 12 Mar 2015 02:11:33 +0000 (22:11 -0400)]
Stop using RPC namespace to unbreak rolling upgrades
This is a temporary patch until we get an Oslo messaging
release that supports Targets with multiple namespaces:
https://review.openstack.org/#/c/163673/
Assaf Muller [Wed, 11 Mar 2015 15:45:00 +0000 (11:45 -0400)]
Fix pylint issue with type VS isinstance in event_observers
New version of pylint started complaining about the unpythonic
usage of type. This patch fixes the issue. Another patch will
propose to pin the version of pylint.
Ihar Hrachyshka [Tue, 10 Mar 2015 21:57:58 +0000 (22:57 +0100)]
Added oslo.log dependency
This is needed for vendor libraries out of tree to start switching their
code to oslo.log logging interface. Hopefully, they will be able to run
against neutron core using neutron.openstack.common.log while executing
oslo_log.log.* calls.
Code Review / openstack-build / neutron-build.git / log