]> review.fuel-infra Code Review - openstack-build/neutron-build.git/log
openstack-build/neutron-build.git
9 years agoFix a problem with openvswitch-agent where it's config file was not being loaded.
gustavo panizzo [Wed, 6 May 2015 06:37:44 +0000 (14:37 +0800)]
Fix a problem with openvswitch-agent where it's config file was not being loaded.

9 years agoFixed gbp.conf
Thomas Goirand [Thu, 30 Apr 2015 21:55:11 +0000 (21:55 +0000)]
Fixed gbp.conf

9 years agoUploading final kilo 2015.1.0
Thomas Goirand [Thu, 30 Apr 2015 21:54:44 +0000 (21:54 +0000)]
Uploading final kilo 2015.1.0

9 years agoMerge tag '2015.1.0' into debian/kilo
Thomas Goirand [Thu, 30 Apr 2015 21:54:06 +0000 (21:54 +0000)]
Merge tag '2015.1.0' into debian/kilo

Neutron 2015.1.0

9 years agoFixed build-depends.
Thomas Goirand [Tue, 28 Apr 2015 06:57:05 +0000 (06:57 +0000)]
Fixed build-depends.

9 years agoNow packaging Kilo rc2.
Thomas Goirand [Tue, 28 Apr 2015 06:54:42 +0000 (06:54 +0000)]
Now packaging Kilo rc2.

9 years agoMerge tag '2015.1_rc2' into debian/kilo
Thomas Goirand [Tue, 28 Apr 2015 06:52:45 +0000 (06:52 +0000)]
Merge tag '2015.1_rc2' into debian/kilo

Neutron kilo-rc2 milestone (2015.1.0rc2)

9 years agoRun radvd as root
Henry Gessau [Mon, 27 Apr 2015 13:59:21 +0000 (09:59 -0400)]
Run radvd as root

During the refactoring of external process management radvd lost
its root privileges.

Closes-bug: 1448813

Change-Id: I84883fe81684afafac9b024282a03f447c8f825a
(cherry picked from commit a5e54338770fc074e01fa88dbf909ee1af1b66b2)

9 years agoDon't resync on DHCP agent setup failure
Kevin Benton [Fri, 24 Apr 2015 07:35:31 +0000 (00:35 -0700)]
Don't resync on DHCP agent setup failure

There are various cases where the DHCP agent will try to
create a DHCP port for a network and there will be a failure.
This has primarily been caused by a lack of available IP addresses
in the allocation pool. Trying to fix all availability corner cases
on the server side will be very difficult due to race conditions between
multiple ports being created, the dhcp_agents_per_network parameter, etc.

This patch just stops the resync attempt on the agent side if a failure
is caused by an IP address generation problem. Future updates to the subnet
will cause another attempt so if the tenant does fix the issue they will
get DHCP service.

Change-Id: I0896730126d6dca13fe9284b4d812cfb081b6218
Closes-Bug: #1447883
(cherry picked from commit db9ac7e0110a0c2ef1b65213317ee8b7f1053ddc)

9 years agoMerge "Restrict subnet create/update to avoid DHCP resync" into stable/kilo
Jenkins [Thu, 23 Apr 2015 14:08:02 +0000 (14:08 +0000)]
Merge "Restrict subnet create/update to avoid DHCP resync" into stable/kilo

9 years agoMerge "Double functional testing timeout to 180s" into stable/kilo
Jenkins [Thu, 23 Apr 2015 14:07:10 +0000 (14:07 +0000)]
Merge "Double functional testing timeout to 180s" into stable/kilo

9 years agoMerge "tests: confirm that _output_hosts_file does not log too often" into stable...
Jenkins [Thu, 23 Apr 2015 13:57:49 +0000 (13:57 +0000)]
Merge "tests: confirm that _output_hosts_file does not log too often" into stable/kilo

9 years agoMerge "Updated from global requirements" into stable/kilo
Jenkins [Thu, 23 Apr 2015 09:48:21 +0000 (09:48 +0000)]
Merge "Updated from global requirements" into stable/kilo

9 years agoMerge "Only update MTU in update code for MTU" into stable/kilo
Jenkins [Thu, 23 Apr 2015 09:48:08 +0000 (09:48 +0000)]
Merge "Only update MTU in update code for MTU" into stable/kilo

9 years agoMerge "Release Import of Translations from Transifex" into stable/kilo
Jenkins [Thu, 23 Apr 2015 09:47:55 +0000 (09:47 +0000)]
Merge "Release Import of Translations from Transifex" into stable/kilo

9 years agotests: confirm that _output_hosts_file does not log too often
Ihar Hrachyshka [Mon, 20 Apr 2015 15:06:38 +0000 (17:06 +0200)]
tests: confirm that _output_hosts_file does not log too often

I3ad7864eeb2f959549ed356a1e34fa18804395cc didn't include any regression unit
tests to validate that the method won't ever log too often again,
reintroducing performance drop in later patches. It didn't play well
with stable backports of the fix, where context was lost when doing the
backport, that left the bug unfixed in stable/juno even though the patch
was merged there [1].

The patch adds an explicit note in the code that suggests not to add new
log messages inside the loop to avoid regression, and a unit test was
added to capture it.

Once the test is merged in master, it will be proposed for stable/juno
inclusion, with additional changes that would fix the regression again.

Related-Bug: #1414218
Change-Id: I5d43021932d6a994638c348eda277dd8337cf041
(cherry picked from commit 3b74095a935f6d2027e6bf04cc4aa21f8a1b46f2)

9 years agoFixed section of the rabbitmq credentials
Thomas Goirand [Thu, 23 Apr 2015 09:10:57 +0000 (11:10 +0200)]
Fixed section of the rabbitmq credentials

9 years agoDouble functional testing timeout to 180s
Maru Newby [Thu, 9 Apr 2015 17:00:57 +0000 (17:00 +0000)]
Double functional testing timeout to 180s

The increase in ovs testing is resulting in job failure due to
timeouts in test_killed_monitor_respawns.  Giving the test more
time to complete should reduce the failure rate.

Change-Id: I2ba9b1eb388bfbbebbd6b0f3edb6d5a5ae0bfead
Closes-Bug: #1442272
(cherry picked from commit 81098620c298394e1a98127ceeba7f297db2d906)

9 years agoAdded missing dnsmasq-neutron.conf in /etc/neutron
Thomas Goirand [Thu, 23 Apr 2015 07:25:44 +0000 (09:25 +0200)]
Added missing dnsmasq-neutron.conf in /etc/neutron

9 years agoRestrict subnet create/update to avoid DHCP resync
watanabe.isao [Wed, 15 Apr 2015 06:48:08 +0000 (15:48 +0900)]
Restrict subnet create/update to avoid DHCP resync

As we know, IPs in subnet CIDR are used for
1) Broadcast port
2) Gateway port
3) DHCP port if enable_dhcp is True, or update to True
4) Others go into allocation_pools
Above 1) to 3) are created by default, which means if CIDR doesn't
have that much of IPs, subnet create/update will cause a DHCP resync.

This fix is to add some restricts to the issue:
A) When subnet create, if enable_dhcp is True, /31 and /32
   cidrs are forbidden for IPv4 subnets while /127 and /128 cidrs are
   forbidden for IPv6 subnets.
B) When subnet update, if enable_dhcp is changing to True and there are no
   more IPs in allocation_pools, the request should be denied.

Change-Id: I2e4a4d5841b9ad908f02b7d0795cba07596c023d
Co-authored-by: Andrew Boik <dboik@cisco.com>
Closes-Bug: #1443798
(cherry picked from commit 0c1f96ad5a6606c1205bd50ea944c3a383892cde)

9 years agoOnly update MTU in update code for MTU
Kevin Benton [Tue, 21 Apr 2015 05:26:22 +0000 (22:26 -0700)]
Only update MTU in update code for MTU

The ML2 create_network_db was re-passing in the entire network
with extensions like vlan_transparency present that was causing
issues in the base update function it was calling.

This corrects the behavior by having it only update the MTU, which
is the only thing it was intending to update in the first place.

Change-Id: I723c5c138e0830de98f6024c7635ec65065e9346
Closes-Bug: #1446784
(cherry picked from commit f85de393c469d1e649a1c1e5ee1b683246442351)

9 years agoUpdated from global requirements
OpenStack Proposal Bot [Thu, 23 Apr 2015 02:15:06 +0000 (02:15 +0000)]
Updated from global requirements

Change-Id: I514c65fac38ef0e534e7401a5f3643b1906adea7

9 years agoMerge "Set IPset hash type to 'net' instead of 'ip'" into stable/kilo
Jenkins [Wed, 22 Apr 2015 19:35:56 +0000 (19:35 +0000)]
Merge "Set IPset hash type to 'net' instead of 'ip'" into stable/kilo

9 years agoMerge "Removed ml2_conf_odl.ini config file" into stable/kilo
Jenkins [Wed, 22 Apr 2015 15:24:18 +0000 (15:24 +0000)]
Merge "Removed ml2_conf_odl.ini config file" into stable/kilo

9 years agoMerge "Revert "Add ipset element and hashsize tunables"" into stable/kilo
Jenkins [Wed, 22 Apr 2015 15:24:04 +0000 (15:24 +0000)]
Merge "Revert "Add ipset element and hashsize tunables"" into stable/kilo

9 years agoMerge "Add Kilo release milestone" into stable/kilo
Jenkins [Wed, 22 Apr 2015 15:17:46 +0000 (15:17 +0000)]
Merge "Add Kilo release milestone" into stable/kilo

9 years agoMerge "ML2: Change port status only when it's bound to the host" into stable/kilo
Jenkins [Wed, 22 Apr 2015 10:19:59 +0000 (10:19 +0000)]
Merge "ML2: Change port status only when it's bound to the host" into stable/kilo

9 years agoMerge "IPv6 SLAAC subnet create should update ports on net" into stable/kilo
Jenkins [Wed, 22 Apr 2015 10:19:47 +0000 (10:19 +0000)]
Merge "IPv6 SLAAC subnet create should update ports on net" into stable/kilo

9 years agoFixed the fix for service_plugins
Thomas Goirand [Tue, 21 Apr 2015 09:10:53 +0000 (11:10 +0200)]
Fixed the fix for service_plugins

9 years agoFixed service_plugins directive.
Thomas Goirand [Tue, 21 Apr 2015 08:05:40 +0000 (10:05 +0200)]
Fixed service_plugins directive.

9 years agoAdded a debian/neutron-server.service.in
Thomas Goirand [Mon, 20 Apr 2015 21:14:44 +0000 (23:14 +0200)]
Added a debian/neutron-server.service.in

9 years agoML2: Change port status only when it's bound to the host
mathieu-rohon [Sat, 7 Mar 2015 12:30:49 +0000 (13:30 +0100)]
ML2: Change port status only when it's bound to the host

Currently, nothing prevents the port status to be changed to BUILD
state when get_device_details() is sent by a host that doesn't own
the port.
In some cases the port might stay in BUILD state.
This could happen during a live-migration, or for multi-hosted ports
such as HA ports.
This commit allows the port status modification only if the port
is bound to the host that is asking for it.

Closes-Bug: #1439857
Closes-Bug: #1438040
Closes-Bug: #1416933

Change-Id: I9b3673f453abbafaaa4f78542fcfebe8dc93f2bb
(cherry picked from commit 9b53b82ce7dad551ebc0f02ff667d5345fb7e139)

9 years agoRelease Import of Translations from Transifex
Andreas Jaeger [Mon, 20 Apr 2015 09:07:37 +0000 (11:07 +0200)]
Release Import of Translations from Transifex

Manual import of Translations from Transifex. This change also removes
all po files that are less than 66 per cent translated since such
partially translated files will not help users.

This updates also recreates all pot (translation source files) to
reflect the state of the repository.

This change needs to be done manually since the automatic import does
not handle the proposed branches and we need to sync with latest
translations.

Change-Id: I1b7bd1773bcd12ab282e77ee0dc41c27846fb66b

9 years agoDo not build-depends on lbaas
Thomas Goirand [Sun, 19 Apr 2015 21:17:05 +0000 (23:17 +0200)]
Do not build-depends on lbaas

9 years agoFixed (build-)depends versions for Jessie.
Thomas Goirand [Sat, 18 Apr 2015 19:55:42 +0000 (21:55 +0200)]
Fixed (build-)depends versions for Jessie.

9 years agoFixes race condition and boosts the scheduling performance
Swaminathan Vasudevan [Wed, 15 Apr 2015 04:34:33 +0000 (21:34 -0700)]
Fixes race condition and boosts the scheduling performance

This patch fixes a race-condition that occurs when the
scheduler tries to check for dvr serviceable ports before
it schedules a router when a subnet is associated with
a router.

Sometimes the dhcp port creation is delayed and so the
router is not scheduled to the l3-agent.

Also it boosts the scheduling performance on dvr-snat
node for scheduling a router.

This patch will provide a work around to fix this race
condition and to boost the scheduling performance
by scheduling a router on a dvr-snat when
dhcp is enabled on the provided subnet, instead of checking
all the available ports on the subnet.

Closes-Bug: #1442494

Change-Id: I089fefdd8535bdc9ed90b3230438ab0bfb6aab4f
(cherry picked from commit c65d3ab6ad4589e6e4a6b488d2eb5d1e4cfee138)

9 years agoSet IPset hash type to 'net' instead of 'ip'
Kevin Benton [Tue, 31 Mar 2015 06:52:56 +0000 (23:52 -0700)]
Set IPset hash type to 'net' instead of 'ip'

The previous hash type was 'ip' and this caused a major
issue with the allowed address pairs extension since it
results in CIDRs being passed to ipset. When the hash type
is 'ip', a CIDR is completely enumerated into all of its
addresses so 10.100.0.0/16 results in ~65k entries. This
meant a single allowed_address_pairs entry could easily
exhaust an entire set.

This patch changes the hash type to 'net', which is designed
to handle a CIDRs as a single entry.

This patch also changes the names of the ipsets because
creating an ipset with different parameters will cause an
error and our ipset manager code isn't robust enough to handle
that at this time. There is another ongoing patch to fix
that but it won't be ready in time.[1]

The related bug was closed by increasing the set limit, which
did alleviate the problem. However, this change would also
address the issue because the gate tests run an allowed address
pairs extension test with the CIDR mentioned above.

1. I59e2e1c090cb95ee1bd14dbb53b6ff2c5e2713fd

Related-Bug: #1439817
Closes-Bug: #1444397
Change-Id: I8177699b157cd3eac46e2f481f47b5d966c49b07
(cherry picked from commit a38b5df5cd3c47672705aad4c30e789ae11ec958)

9 years agoRevert "Add ipset element and hashsize tunables"
Kevin Benton [Tue, 31 Mar 2015 15:53:56 +0000 (08:53 -0700)]
Revert "Add ipset element and hashsize tunables"

This reverts commit b5b919a7a3569ccb93c3d7d523c1edfaeddb7cb9.

The current ipset manager code isn't robust enough to handle
ipsets that already exist with different parameters. This reverts
the ability to change the parameters so we don't break upgrades
to Kilo.

Conflicts:
neutron/agent/linux/ipset_manager.py
neutron/tests/unit/agent/linux/test_ipset_manager.py

Change-Id: I538714df52424f0502cb75daea310517d1142c42
Closes-Bug: #1444201
(cherry picked from commit 03be14a569d240865dabff8b4c30385abf1dbe62)

9 years agoUpdate .gitreview to point to stable/kilo
Kevin Benton [Thu, 16 Apr 2015 23:27:38 +0000 (16:27 -0700)]
Update .gitreview to point to stable/kilo

This is the stable/kilo branch. When people make changes
here it's highly likely that they want to propose them
to stable/kilo on gerrit.

Change-Id: Ie61a9f0c0b0b4896da33a201e42b1c4bc4bae49b

9 years agoAdd Kilo release milestone
Henry Gessau [Thu, 16 Apr 2015 17:38:46 +0000 (13:38 -0400)]
Add Kilo release milestone

Change-Id: Id7d969c92b7c757b766760681357ac13c8079ca3

9 years agoIPv6 SLAAC subnet create should update ports on net
Dane LeBlanc [Thu, 9 Apr 2015 14:32:33 +0000 (10:32 -0400)]
IPv6 SLAAC subnet create should update ports on net

If ports are first created on a network, and then an IPv6 SLAAC
or DHCPv6-stateless subnet is created on that network, then the
ports created prior to the subnet create are not getting
automatically updated (associated) with addresses for the
SLAAC/DHCPv6-stateless subnet, as required.

Change-Id: I88d04a13ce5b8ed4c88eac734e589e8a90e986a0
Closes-Bug: 1427474
Closes-Bug: 1441382
Closes-Bug: 1440183
(cherry picked from commit bd1044ba0e9d7d0f4752c891ac340b115f0019c4)

9 years agoRemoved ml2_conf_odl.ini config file
Ihar Hrachyshka [Fri, 10 Apr 2015 13:07:33 +0000 (15:07 +0200)]
Removed ml2_conf_odl.ini config file

The file is already packaged into decomposed networking-odl repo [1].

[1]: https://git.openstack.org/cgit/stackforge/networking-odl/tree/etc/neutron/plugins/ml2/ml2_conf_odl.ini

Closes-Bug: #1442615
Change-Id: Ic280454190aab4e3b881cde15a882808b652861e
(cherry picked from commit b3334eca0ae9f9c64ccd646035e69081f669e3e4)

9 years agoAdded a neutron-plugin-linuxbridge-agent
Thomas Goirand [Wed, 15 Apr 2015 16:17:58 +0000 (18:17 +0200)]
Added a neutron-plugin-linuxbridge-agent

9 years agoMerge branch 'debian/kilo' of git+ssh://git.debian.org/git/openstack/neutron into...
Thomas Goirand [Wed, 15 Apr 2015 16:17:15 +0000 (18:17 +0200)]
Merge branch 'debian/kilo' of git+ssh://git.debian.org/git/openstack/neutron into debian/kilo

Conflicts:
debian/control

9 years agoUsing the build dir for running tests.
Thomas Goirand [Wed, 15 Apr 2015 16:07:27 +0000 (18:07 +0200)]
Using the build dir for running tests.

9 years agoFixed (build-)depends for this release.
Thomas Goirand [Wed, 15 Apr 2015 14:52:18 +0000 (16:52 +0200)]
Fixed (build-)depends for this release.

9 years agoDelete obsolete patches.
Thomas Goirand [Wed, 15 Apr 2015 14:04:00 +0000 (16:04 +0200)]
Delete obsolete patches.

9 years agoDo not attempt to "fix" alembic migrations.
Thomas Goirand [Wed, 15 Apr 2015 14:02:41 +0000 (16:02 +0200)]
Do not attempt to "fix" alembic migrations.

9 years agoMerge tag '2015.1_rc1' into debian/kilo
Thomas Goirand [Wed, 15 Apr 2015 14:02:15 +0000 (16:02 +0200)]
Merge tag '2015.1_rc1' into debian/kilo

Neutron kilo-rc1 milestone (2015.1.0rc1)

9 years agoNow packaging Kilo rc1.
Thomas Goirand [Wed, 15 Apr 2015 14:02:06 +0000 (16:02 +0200)]
Now packaging Kilo rc1.

9 years agoBreaks and Replaces for the rename of neutron openvswitch and
gustavo panizzo [Fri, 10 Apr 2015 03:10:00 +0000 (11:10 +0800)]
Breaks and Replaces for the rename of neutron openvswitch and
linuxbridge agents.

9 years agorename neutron-plugin-[openvswitch|linuxbridge]-agent to
gustavo panizzo [Fri, 10 Apr 2015 02:28:18 +0000 (10:28 +0800)]
rename neutron-plugin-[openvswitch|linuxbridge]-agent to
neutron-[openvswitch|linuxbridge]-agent

9 years agoRevert "add -plugin- to the names of the lb and ovs agents"
gustavo panizzo [Thu, 9 Apr 2015 11:06:11 +0000 (19:06 +0800)]
Revert "add -plugin- to the names of the lb and ovs agents"

This reverts commit f17e5be3b8243764793bdc27e129c111e655f9f0.

9 years agoadd -plugin- to the names of the lb and ovs agents
gustavo panizzo [Thu, 9 Apr 2015 10:26:35 +0000 (18:26 +0800)]
add -plugin- to the names of the lb and ovs agents

9 years agodocument the changes on init.d scripts for ovs and lb agents
gustavo panizzo [Thu, 9 Apr 2015 09:30:12 +0000 (17:30 +0800)]
document the changes on init.d scripts for ovs and lb agents

9 years agodo not parse neutron.conf to know the core_plugin name in ovs and lb
gustavo panizzo [Wed, 8 Apr 2015 08:44:12 +0000 (16:44 +0800)]
do not parse neutron.conf to know the core_plugin name in ovs and lb
agent.

OVS and LB do not work as core_plugin since Juno, they need ML2 to work.

9 years agoMerge "Add simple ARP spoofing protection"
Jenkins [Thu, 9 Apr 2015 01:21:53 +0000 (01:21 +0000)]
Merge "Add simple ARP spoofing protection"

9 years agoMerge "Add missing config parameters in neutron.conf"
Jenkins [Wed, 8 Apr 2015 22:42:45 +0000 (22:42 +0000)]
Merge "Add missing config parameters in neutron.conf"

9 years agoMerge "Re-use context session in ML2 DB get_port_binding_host"
Jenkins [Wed, 8 Apr 2015 22:40:49 +0000 (22:40 +0000)]
Merge "Re-use context session in ML2 DB get_port_binding_host"

9 years agoMerge "Revert "IPv6 SLAAC subnet create should update ports on net""
Jenkins [Wed, 8 Apr 2015 20:39:59 +0000 (20:39 +0000)]
Merge "Revert "IPv6 SLAAC subnet create should update ports on net""

9 years agoMerge "Moving VLAN Transparency support from core to extension"
Jenkins [Wed, 8 Apr 2015 20:30:01 +0000 (20:30 +0000)]
Merge "Moving VLAN Transparency support from core to extension"

9 years agoRevert "IPv6 SLAAC subnet create should update ports on net"
armando-migliaccio [Wed, 8 Apr 2015 17:57:13 +0000 (10:57 -0700)]
Revert "IPv6 SLAAC subnet create should update ports on net"

This reverts commit 81f4469b620ec221f53d3ffb4d00b90896dc5ce1.

Change-Id: I63a392fccda29ceff3e91c0a4de741d263bd0e8e
Related-bug: #1441382
Related-bug: #1440183

9 years agoAdd missing config parameters in neutron.conf
Edgar Magana [Tue, 7 Apr 2015 05:57:06 +0000 (22:57 -0700)]
Add missing config parameters in neutron.conf

Include all missing configuration parameters already
integrated in Neutron code.

Change-Id: Iefa344a2f9ec2c74f6314e7c783ff3b213d76ea3
Closes-bug: #1438329

9 years agoMerge "Return from check_ports_exist_on_l3agent if no subnet found"
Jenkins [Wed, 8 Apr 2015 16:11:05 +0000 (16:11 +0000)]
Merge "Return from check_ports_exist_on_l3agent if no subnet found"

9 years agoMerge "Add ipset element and hashsize tunables"
Jenkins [Wed, 8 Apr 2015 15:57:10 +0000 (15:57 +0000)]
Merge "Add ipset element and hashsize tunables"

9 years agoMoving VLAN Transparency support from core to extension
Pritesh Kothari [Wed, 25 Mar 2015 18:34:05 +0000 (11:34 -0700)]
Moving VLAN Transparency support from core to extension

* Moving VLAN Transparency support from core to extension
* Remove the older unit tests and add new corresponding ones

DocImpact
Closes-Bug: #1434667

Change-Id: Ic551475ed7b64aad9627a57abb0df41acc19bfc1

9 years agoRe-use context session in ML2 DB get_port_binding_host
Dane LeBlanc [Sat, 4 Apr 2015 22:50:36 +0000 (18:50 -0400)]
Re-use context session in ML2 DB get_port_binding_host

This patch modifies ML2 DB get_port_binding_host method so that it
reuses the existing context session to do the database query
rather than creating a new database session.

Note that there are other methods in ML2 DB that do not re-use
the caller's session (get_port_from_device_mac() and
get_sg_ids_grouped_by_port()). These will be modified using
a separate bug (https://bugs.launchpad.net/neutron/+bug/1441205).
Change-Id: I8aafb0a70f40f9306ccc366e5db6860c92c48cce
Closes-Bug: #1440183

9 years agoMerge "linuxbridge UT: Fix a regression of the recent ip_lib change"
Jenkins [Wed, 8 Apr 2015 15:09:25 +0000 (15:09 +0000)]
Merge "linuxbridge UT: Fix a regression of the recent ip_lib change"

9 years agoMerge "Remove tests from HA routers test framework"
Jenkins [Wed, 8 Apr 2015 05:37:22 +0000 (05:37 +0000)]
Merge "Remove tests from HA routers test framework"

9 years agoMerge "Fix intermittent UT failures in test_utils"
Jenkins [Wed, 8 Apr 2015 04:40:45 +0000 (04:40 +0000)]
Merge "Fix intermittent UT failures in test_utils"

9 years agoFix intermittent UT failures in test_utils
armando-migliaccio [Tue, 7 Apr 2015 22:37:59 +0000 (15:37 -0700)]
Fix intermittent UT failures in test_utils

Change eba4c2941ee introduced these tests. However they are not that useful as they
simply mimick the code, without really ensuring that the behavior is expected, so
they provide negative value ([1]), plus, they fail randomly.

This patch removes them in favor of a more useful functional check.

[1] http://googletesting.blogspot.com/2015/01/testing-on-toilet-change-detector-tests.html

Closes-bug: #1441347

Change-Id: I8a321995295deef7f6d30be303486be491e2771f

9 years agoMerge "Enhance TESTING.rst"
Jenkins [Tue, 7 Apr 2015 22:19:56 +0000 (22:19 +0000)]
Merge "Enhance TESTING.rst"

9 years agoMerge "Remove check for bash usage"
Jenkins [Tue, 7 Apr 2015 21:26:04 +0000 (21:26 +0000)]
Merge "Remove check for bash usage"

9 years agoMerge "Stop using deprecated DEFAULT group for lock_path"
Jenkins [Tue, 7 Apr 2015 18:38:52 +0000 (18:38 +0000)]
Merge "Stop using deprecated DEFAULT group for lock_path"

9 years agoMerge "Fix dynamic arp populate error for dvr routers"
Jenkins [Tue, 7 Apr 2015 18:28:26 +0000 (18:28 +0000)]
Merge "Fix dynamic arp populate error for dvr routers"

9 years agoEnhance TESTING.rst
Maru Newby [Tue, 24 Mar 2015 19:45:46 +0000 (19:45 +0000)]
Enhance TESTING.rst

Add detail about api testing and provide better visual separation
between the different types of testing.

The current testing guidelines are mainly about running tests, and
this change does little to fix that.  The intention is to add detail
about writing tests in subsequent changes.

Change-Id: I39d0439c91e5c6edb1d48d4da310443c99fb6d9e

9 years agoRemove check for bash usage
Maru Newby [Tue, 7 Apr 2015 15:00:25 +0000 (15:00 +0000)]
Remove check for bash usage

Arbitrarily restricting ourselves from using bash because developers on
platforms like netbsd don't want to install bash from ports doesn't
make sense.  Any non-trivial shell script is likely to use features
like arrays or string manipulation that are poorly supported (if at
all) by sh, and the continued bumping of the number of expected bash
scripts is an indication that the check is not serving its purpose
anyway.

Along with removing the check, all shebang references to /bin/bash
have been replaced with /usr/bin/env bash in an attempt to be more
compatible across different hosts.

Change-Id: Ief72dc380cc88af38959c330897e2c127e33c332
Closes-Bug: #1440824

9 years agoReturn from check_ports_exist_on_l3agent if no subnet found
Ed Bak [Mon, 9 Feb 2015 23:13:18 +0000 (23:13 +0000)]
Return from check_ports_exist_on_l3agent if no subnet found

The call to get_subnet_ids_on_router can return an empty list.
If the subnet_ids list is empty, the subsequent call to get
the ports on a subnet returns all ports.  If this occurs
when doing a remove_router_interface, the performance
of a remove_router_interface degrades significantly.  This change
returns immediately from check_ports_exist_on_l3agents if no
subnet is found.  A new unit test has been added to cover
the specific case of returning immediately without calling
get_ports when a remove_router_interface operation is performed.

Change-Id: I247d3bae152ab4f8ab7e00bd24d878eb08dca1ba
Closes-Bug: #1420032
Depends-On: I15bbf16fd4378c6431e9da8942d0968e7a012a91

9 years agoRemove tests from HA routers test framework
Assaf Muller [Tue, 7 Apr 2015 03:56:15 +0000 (23:56 -0400)]
Remove tests from HA routers test framework

The framework class should not contain tests. Running the module
was running an additional 9 tests that should not have been run.

Change-Id: Iabc6367e8bfda18e395d1a19809b07507200003d

9 years agolinuxbridge UT: Fix a regression of the recent ip_lib change
YAMAMOTO Takashi [Wed, 18 Mar 2015 04:27:15 +0000 (13:27 +0900)]
linuxbridge UT: Fix a regression of the recent ip_lib change

A recently merged change, I07d1d297f07857d216649cccf717896574aac301,
changed IPWrapper.get_devices to use /sys instead of executing ip command.
Unfortunately it broke linuxbridge unit tests, which seems to assume that
mocking utils.execute is enough in some places.  This commit fixes the
regression.

Closes-Bug: #1433417
Related-Bug: #1374663
Change-Id: I9570abe703b438a3fc358f747e25d023934d1ffd

9 years agoFix dynamic arp populate error for dvr routers
Swaminathan Vasudevan [Fri, 3 Apr 2015 00:25:39 +0000 (17:25 -0700)]
Fix dynamic arp populate error for dvr routers

Recent refactor to the L3 Agent have introduced
this problem. When we create a VM after we
attach an interface to a router or when we add
an interface with an existing VM to a router, in
both cases the arp entries for the dvr serviced
ports are not getting populated in the Router
Namespace.

Closes-Bug: #1438969

Change-Id: I4a82e2435d176f3d9336d7f0dab9726c063840b9
Co-authored-by: Armando Migliaccio <armamig@gmail.com>
9 years agoReorganize plugin test modules
Maru Newby [Mon, 6 Apr 2015 21:53:39 +0000 (21:53 +0000)]
Reorganize plugin test modules

This change moves plugin test modules to conform to the new rules on
unit test tree structure (see TESTING.rst).

Vendor plugin paths continue to be ignored, and unit test modules that
test features instead of modules are also ignored pending their
removal to the functional test tree.

Change-Id: I482c377ca72ffd58692ad84bd9692356513e4c98
Closes-Bug: #1440834

9 years agoMerge open source plugin test code modules
Maru Newby [Mon, 6 Apr 2015 21:51:23 +0000 (21:51 +0000)]
Merge open source plugin test code modules

The unit test reorg is about moving files around so a test module is
clearly associated with the code module it targets, but the test
modules in this change needed to be manually merged because they both
targeted the same module.

Change-Id: I80f4b97fadd318896e7fa4e7e7e939f924127b2a
Partial-Bug: #1440834

9 years agoReorganize unit test tree
Maru Newby [Sat, 4 Apr 2015 00:22:05 +0000 (00:22 +0000)]
Reorganize unit test tree

This change ensures that the structure of the unit test tree matches
that of the code tree to make it obvious where to find tests for a
given module.  A check is added to the pep8 job to protect against
regressions.

The plugin test paths are relocated to neutron/tests/unit/plugins
but are otherwise ignored for now.

Change-Id: If307593259139171be21a71c58e3a34bf148cc7f
Partial-Bug: #1440834

9 years agoMerge "Fix docstring for l3_dvr_db.dvr_vmarp_table_update"
Jenkins [Mon, 6 Apr 2015 22:33:45 +0000 (22:33 +0000)]
Merge "Fix docstring for l3_dvr_db.dvr_vmarp_table_update"

9 years agoMerge "Allow metadata proxy running with nobody user/group"
Jenkins [Mon, 6 Apr 2015 21:12:33 +0000 (21:12 +0000)]
Merge "Allow metadata proxy running with nobody user/group"

9 years agoAdd ipset element and hashsize tunables
Brian Haley [Fri, 3 Apr 2015 01:11:06 +0000 (21:11 -0400)]
Add ipset element and hashsize tunables

Recently, these messages have been noticed in both tempest
logs, as well as reported by downstream users syslog:

  Set IPv4915d358d-2c5b-43b5-9862 is full, maxelem 65536 reached

So the default of 64K is not sufficient enough.

This change adds two config options to control both the number
of elements as well as the hashsize, since they should be
tuned together for best performance.  Slightly different
formats were required for 'ipset create' and 'ipset restore'.

The default values for these are now set to 131072 (maxelem) and
2048 (hashsize), which is an increase over their typical default values
of 65536/1024 (respectively), in order to fix the errors seen in
the tempest tests.

DocImpact

Change-Id: Ic0b5b38a840e737dc6be938230f4052974c8620f
Closes-bug: #1439817

9 years agoAllow metadata proxy running with nobody user/group
Cedric Brandily [Tue, 17 Mar 2015 15:20:07 +0000 (15:20 +0000)]
Allow metadata proxy running with nobody user/group

Currently metadata proxy cannot run with nobody user/group as metadata
proxy requires to connect to metadata_proxy_socket when queried.

This change allows to run metadata proxy with nobody user/group by
allowing to choose the metadata_proxy_socket mode with the new option
metadata_proxy_socket_mode (4 choices) in order to adapt socket
permissions to metadata proxy user/group.

This change refactors also where options are defined to enable
metadata_proxy_user/group options in the metadata agent.

In practice:
* if metadata_proxy_user is agent effective user or root, then:
  * metadata proxy is allowed to use rootwrap (unsecure)
  * set metadata_proxy_socket_mode = user (0o644)
* else if metadata_proxy_group is agent effective group, then:
  * metadata proxy is not allowed to use rootwrap (secure)
  * set metadata_proxy_socket_mode = group (0o664)
  * set metadata_proxy_log_watch = false
* else:
  * metadata proxy has lowest permissions (securest) but metadata proxy
    socket can be opened by everyone
  * set metadata_proxy_socket_mode = all (0o666)
  * set metadata_proxy_log_watch = false

An alternative is to set metadata_proxy_socket_mode = deduce, in such
case metadata agent uses previous rules to choose the correct mode.

DocImpact
Closes-Bug: #1427228
Change-Id: I235a0cc4f0cbd55ae4ec1570daf2ebbb6a72441d

9 years agoSkip example retargetable functional test
Maru Newby [Mon, 6 Apr 2015 16:28:00 +0000 (16:28 +0000)]
Skip example retargetable functional test

The example retargetable test that previously ran as part of the
functional suite is now skipped due to the fullstack example's db
fixture usage causing the test to fail if it the fullstack example
runs first on the same worker.

Change-Id: I0a34f9ba04c53a4291698be819070c66009c8b4a
Related-Bug: #1440797

9 years agoMerge "Prepare for unit test reorg"
Jenkins [Mon, 6 Apr 2015 13:37:14 +0000 (13:37 +0000)]
Merge "Prepare for unit test reorg"

9 years agoMerge "Store and log correct exception info"
Jenkins [Sun, 5 Apr 2015 05:43:37 +0000 (05:43 +0000)]
Merge "Store and log correct exception info"

9 years agoMerge "Always run dnsmasq as root"
Jenkins [Sun, 5 Apr 2015 05:40:37 +0000 (05:40 +0000)]
Merge "Always run dnsmasq as root"

9 years agoMerge "Move network MTU from core REST API to extension API"
Jenkins [Sat, 4 Apr 2015 20:13:55 +0000 (20:13 +0000)]
Merge "Move network MTU from core REST API to extension API"

9 years agoMerge "Refactoring cleanup for L3 agent callbacks"
Jenkins [Sat, 4 Apr 2015 04:55:20 +0000 (04:55 +0000)]
Merge "Refactoring cleanup for L3 agent callbacks"

9 years agoPrepare for unit test reorg
Maru Newby [Fri, 3 Apr 2015 23:42:31 +0000 (23:42 +0000)]
Prepare for unit test reorg

The unit test reorg is about moving files around so a test module is
clearly associated with the code module it targets, but the test
modules in this change needed to be manually merged because they both
targeted the same module.

test_api_v2 is also updated to use the path of neutron/tests/base.py
as the root of path to test implementations of extensions.

Change-Id: I432b84339e51c26ef0aa26d44e29b5a3311626ad
Implements: bp/reorganize-unit-test-tree

9 years agoMerge "Add API tests for subnet-create with subnetpool"
Jenkins [Fri, 3 Apr 2015 23:34:24 +0000 (23:34 +0000)]
Merge "Add API tests for subnet-create with subnetpool"

9 years agoMerge "Support multiple IPv6 prefixes on internal router ports"
Jenkins [Fri, 3 Apr 2015 21:49:21 +0000 (21:49 +0000)]
Merge "Support multiple IPv6 prefixes on internal router ports"

9 years agoMerge "Make L3 agent honor periodic_interval setting"
Jenkins [Fri, 3 Apr 2015 21:43:51 +0000 (21:43 +0000)]
Merge "Make L3 agent honor periodic_interval setting"

9 years agoRemove orphaned nuage unit test module
Maru Newby [Fri, 3 Apr 2015 17:26:33 +0000 (17:26 +0000)]
Remove orphaned nuage unit test module

Change I6d02df85c7a2c307ad11442d0afdd50c64210af4 implemented the
plugin decomp for nuage but one of the unit tests was missed.

Change-Id: I37e1b3f6645b5f7730218d5ef08ca28f72b91883

9 years agoMerge "IPv6 SLAAC subnet create should update ports on net"
Jenkins [Fri, 3 Apr 2015 16:46:02 +0000 (16:46 +0000)]
Merge "IPv6 SLAAC subnet create should update ports on net"