apex-tripleo-heat-templates.git
8 years agoAdd constraints and timeouts from file in single step
Giulio Fidente [Wed, 18 Nov 2015 21:23:26 +0000 (22:23 +0100)]
Add constraints and timeouts from file in single step

To avoid pcmk reconfiguring the resources on each config change,
we want to apply the constraints and timeouts from file.

We also *do not* want to alter the timeouts for a few ocf resources
which are rabbitmq, neutron-netns-cleanup and neutron-ovs-cleanup

Change-Id: I6875f19e1f34f0fdcf0928421f49b61d857ca7c8
Co-Authored-By: Andrew Beekhof <abeekhof@redhat.com>
8 years agoMerge "Implement Neutron enable_isolated_metadata parameters"
Jenkins [Wed, 18 Nov 2015 11:24:45 +0000 (11:24 +0000)]
Merge "Implement Neutron enable_isolated_metadata parameters"

8 years agoMerge "Verify galera is sync'd in yum_update.sh"
Jenkins [Wed, 18 Nov 2015 10:53:59 +0000 (10:53 +0000)]
Merge "Verify galera is sync'd in yum_update.sh"

8 years agoMerge "neutron: enable nova-event-callback by default"
Jenkins [Tue, 17 Nov 2015 20:13:41 +0000 (20:13 +0000)]
Merge "neutron: enable nova-event-callback by default"

8 years agoMerge "Pass ceph::pool arguments when calling class"
Jenkins [Tue, 17 Nov 2015 17:41:28 +0000 (17:41 +0000)]
Merge "Pass ceph::pool arguments when calling class"

8 years agoneutron: enable nova-event-callback by default
Emilien Macchi [Mon, 21 Sep 2015 17:46:02 +0000 (13:46 -0400)]
neutron: enable nova-event-callback by default

* Add NovaApiVirtualIP string parameter.
* Compute nova_url and nova_admin_auth_url parameters.
* Configure in Hiera neutron::server::notifications::* parameters.
* non-ha: include ::neutron::server::notifications
* ha: include ::neutron::server::notifications and create orchestration
* Set vif_plugging_is_fatal to True so we actually fail if Neutron is not
  able to create the VIF during Nova server creation workflow.

Depends-On: I21dc10396e92906eab4651c318aa2ee62a8e03c7
Change-Id: I02e41f87404e0030d488476680af2f6d45af94ff

8 years agoImplement Neutron enable_isolated_metadata parameters
Cyril Lopez [Tue, 20 Oct 2015 07:48:49 +0000 (09:48 +0200)]
Implement Neutron enable_isolated_metadata parameters

* Use the parameter in Puppet configuration (Hiera) to configure neutron

BZ-1273303

Change-Id: Ic5a7a1f13fd2bc800cadc3a78b1daadbc0394787
Signed-off-by: Cyril Lopez <cylopez@redhat.com>
8 years agoVerify galera is sync'd in yum_update.sh
James Slagle [Wed, 11 Nov 2015 20:54:25 +0000 (15:54 -0500)]
Verify galera is sync'd in yum_update.sh

When the cluster is brought back online after a yum update in
yum_update.sh, we should verify that galera is fully sync'd before
moving on. This ensures the sync is complete before moving on to update
any other nodes in the cluster.

Change-Id: Ie8fc2c5d5214deacea94ca658ac75359b318ced1

8 years agoMerge "Add support for enabling L2 population in Neutron"
Jenkins [Tue, 17 Nov 2015 15:40:44 +0000 (15:40 +0000)]
Merge "Add support for enabling L2 population in Neutron"

8 years agoMerge "Add simple linux bridge as network environment"
Jenkins [Tue, 17 Nov 2015 12:47:20 +0000 (12:47 +0000)]
Merge "Add simple linux bridge as network environment"

8 years agoAdd simple linux bridge as network environment
Jaume Devesa [Tue, 10 Nov 2015 10:45:22 +0000 (10:45 +0000)]
Add simple linux bridge as network environment

Create a bridge for the overcloud services using linux bridge instead of
openvswitch. Some SDNs may be incompatible with openvswitch datapath.

Change-Id: I873368e74ddfd95bf5c6e1f88cec33ba011e09dd

8 years agoAdd support for enabling L2 population in Neutron
Brent Eagles [Thu, 12 Nov 2015 14:56:53 +0000 (11:26 -0330)]
Add support for enabling L2 population in Neutron

This change adds support for enabling/disabling L2 population in
Neutron agents. It currently defaults to false.

Change-Id: I3dd19feb4acb1046bc560b35e5a7a111364ea0d7

8 years agoMerge "Add environment for isolated networks without tunneling VLAN"
Jenkins [Mon, 16 Nov 2015 14:08:12 +0000 (14:08 +0000)]
Merge "Add environment for isolated networks without tunneling VLAN"

8 years agoMerge "Support network isolation without external nets"
Jenkins [Mon, 16 Nov 2015 14:05:41 +0000 (14:05 +0000)]
Merge "Support network isolation without external nets"

8 years agoMerge "Add the GlanceRegistry and Horizon endpoints to EndpointMap"
Jenkins [Mon, 16 Nov 2015 14:01:53 +0000 (14:01 +0000)]
Merge "Add the GlanceRegistry and Horizon endpoints to EndpointMap"

8 years agoMerge "Expose the IP of the Endpoints"
Jenkins [Mon, 16 Nov 2015 14:00:50 +0000 (14:00 +0000)]
Merge "Expose the IP of the Endpoints"

8 years agoMerge "Make CloudName available for Endpoints"
Jenkins [Mon, 16 Nov 2015 13:59:19 +0000 (13:59 +0000)]
Merge "Make CloudName available for Endpoints"

8 years agoMerge "Bump further the stop/start timeout for pcmk/systemd services"
Jenkins [Mon, 16 Nov 2015 10:17:43 +0000 (10:17 +0000)]
Merge "Bump further the stop/start timeout for pcmk/systemd services"

8 years agoMerge "Set start/stop pacemaker resource timeouts for updates"
Jenkins [Mon, 16 Nov 2015 09:49:58 +0000 (09:49 +0000)]
Merge "Set start/stop pacemaker resource timeouts for updates"

8 years agoMerge "Add missing constraints in yum_update.sh"
Jenkins [Mon, 16 Nov 2015 09:47:55 +0000 (09:47 +0000)]
Merge "Add missing constraints in yum_update.sh"

8 years agoMerge "honor the rabbit user and password provided"
Jenkins [Fri, 13 Nov 2015 21:43:51 +0000 (21:43 +0000)]
Merge "honor the rabbit user and password provided"

8 years agoMerge "Fix cinder error when CinderNfsMountOptions option is absent"
Jenkins [Fri, 13 Nov 2015 18:46:51 +0000 (18:46 +0000)]
Merge "Fix cinder error when CinderNfsMountOptions option is absent"

8 years agoPass ceph::pool arguments when calling class
Giulio Fidente [Wed, 11 Nov 2015 10:51:23 +0000 (11:51 +0100)]
Pass ceph::pool arguments when calling class

Pass the ceph::pool properties as arguments to the class call
instead of setting them as class defaults.

Ceph recommends max 32 PGs and min 4 PGs per OSD so this change
also lowers the defaults to 32 which works with 1 OSD, suits well
a scenario with 3 OSDs and is easy to customize in the static
hiera if more than 8 OSDs are deployed.

More info at: https://bugzilla.redhat.com/show_bug.cgi?id=1252546

Change-Id: Ifed11d1857900b2251dfdf69d6b6f168150e6330

8 years agoMerge "Refacter Endpoints into EndpointMap"
Jenkins [Fri, 13 Nov 2015 17:53:30 +0000 (17:53 +0000)]
Merge "Refacter Endpoints into EndpointMap"

8 years agoMerge "Add DeployIdentifier overcloud parameter"
Jenkins [Fri, 13 Nov 2015 17:26:34 +0000 (17:26 +0000)]
Merge "Add DeployIdentifier overcloud parameter"

8 years agoFix cinder error when CinderNfsMountOptions option is absent
Pierre Blanc [Fri, 13 Nov 2015 16:28:54 +0000 (11:28 -0500)]
Fix cinder error when CinderNfsMountOptions option is absent

When I deploy director with NFS backend for cinder,
sometimes I don't need nfs mount options.
If I choose to omit this option, or if the option
is defined to '', the deployment fails.

This patch add just a default value for this option.

Change-Id: Idf708aaecebd5c6db14f48ad2a53d6c2453be5ee
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1281870

8 years agoSet start/stop pacemaker resource timeouts for updates
Jiri Stransky [Fri, 13 Nov 2015 14:12:54 +0000 (15:12 +0100)]
Set start/stop pacemaker resource timeouts for updates

This matches change I6fc18f1ad876c5a25723710a3b20d8ec9519dcba, but we
need it to set it before attempting the cluster stop - yum update -
cluster start cycle, to make sure this cycle doesn't hit the low timeout
limits.

This can be removed once updates from deployments made prior to
I6fc18f1ad876c5a25723710a3b20d8ec9519dcba are no longer supported.

Change-Id: I587136d8d045d213875c657ea5a405074f80c8ad

8 years agoBump further the stop/start timeout for pcmk/systemd services
Giulio Fidente [Fri, 13 Nov 2015 14:00:30 +0000 (15:00 +0100)]
Bump further the stop/start timeout for pcmk/systemd services

This bumps further up the stop/start timeout for the pcmk/systemd
services so that it matches the 100s default set in future pcmk
versions [1].

1. https://github.com/ClusterLabs/pacemaker/commit/17d65e9f44061a4fa14a9cddd6edc403b2d6d2b3

Change-Id: I6fc18f1ad876c5a25723710a3b20d8ec9519dcba

8 years agoMerge "Pin docker version for atomic at 1.8.2"
Jenkins [Thu, 12 Nov 2015 15:57:14 +0000 (15:57 +0000)]
Merge "Pin docker version for atomic at 1.8.2"

8 years agoMerge "Set default start/stop timeout for pcmk services to 95s"
Jenkins [Thu, 12 Nov 2015 11:02:55 +0000 (11:02 +0000)]
Merge "Set default start/stop timeout for pcmk services to 95s"

8 years agoMerge "Change default bond-mode"
Jenkins [Thu, 12 Nov 2015 10:24:46 +0000 (10:24 +0000)]
Merge "Change default bond-mode"

8 years agoAdd DeployIdentifier overcloud parameter
Dan Prince [Wed, 11 Nov 2015 22:46:50 +0000 (17:46 -0500)]
Add DeployIdentifier overcloud parameter

We've heard from end users that it is confusing that puppet
isn't re-executed on a heat stack-update.

This patch adds a new DeployIdentifier parameter which
we can set via client tooling (tripleoclient) to a unique
value so that on each heat stack-update we always execute
all of our configuration deployments.

Change-Id: Ic352ddd30807dc378e5e7b6c396bc53f5d6d5622
Related-bug: #1505430

8 years agoChange default bond-mode
Joe Talerico [Thu, 22 Oct 2015 17:47:04 +0000 (13:47 -0400)]
Change default bond-mode

The default balance-tcp is causing issues with deployments.
Defaulting to active-backup.

After ~ 100 guests (total) connectivity to each guest would become spotty
(simple pings would fail, then become successful.) In /var/log/messages
we saw :
"overcloud-controller-1 kernel: openvswitch: ovs-system: deferred action
limit reached, drop recirc action"

For more details, refer to this link:
http://openvswitch.org/pipermail/discuss/2015-October/019168.html

Change-Id: Ia0f2592a289e13472b98d97057cd516c5048fe59

8 years agoAdd missing constraints in yum_update.sh
James Slagle [Wed, 11 Nov 2015 20:46:52 +0000 (15:46 -0500)]
Add missing constraints in yum_update.sh

Some missing pacemaker constraints were added in the following commits:

https://review.openstack.org/#/c/219770/
https://review.openstack.org/#/c/219665/
https://review.openstack.org/#/c/218931/
https://review.openstack.org/#/c/218930/

Overclouds that were deployed prior to these constraints being added to
tripleo-heat-templates still have the constraints missing. During an
update, stopping and starting the cluster can fail without these
constraints in place.  As a workaround, conditionally add these
contraints in yum_update.sh so that we're sure they're always present
before updating.

Change-Id: Id46c85dbbe5e85d362279661091b17ce1b697fe0

8 years agoMerge "Allow customization of Ceph client user"
Jenkins [Wed, 11 Nov 2015 15:29:28 +0000 (15:29 +0000)]
Merge "Allow customization of Ceph client user"

8 years agoMerge "Allow a user to specify a comma separated list of ntp servers"
Jenkins [Wed, 11 Nov 2015 10:44:05 +0000 (10:44 +0000)]
Merge "Allow a user to specify a comma separated list of ntp servers"

8 years agoAdd the GlanceRegistry and Horizon endpoints to EndpointMap
Mark Chappell [Tue, 3 Nov 2015 16:15:46 +0000 (17:15 +0100)]
Add the GlanceRegistry and Horizon endpoints to EndpointMap

We need to pass details of the Glance Registry and public Horizon
endpoints to the load balancers so add them to the EndpointMap

Change-Id: Ia6261223e7701734f47ce48471c86f690ba3dcd5

8 years agoExpose the IP of the Endpoints
Mark Chappell [Tue, 3 Nov 2015 16:14:04 +0000 (17:14 +0100)]
Expose the IP of the Endpoints

We expose all of the other parameters, so expose the IP too for
consistency

Change-Id: I5c31befde51e398318c7b8c744310212288ad892

8 years agoMake CloudName available for Endpoints
Mark Chappell [Wed, 28 Oct 2015 14:30:41 +0000 (15:30 +0100)]
Make CloudName available for Endpoints

CloudName is the DNS name for the public VIP this means we will likely
want it available for use in the endpoint hostnames, rather than people
needing to copy and paste the same hostname

Change-Id: Ic6d708b083244442195eee890de91bbc7e133ec2

8 years agoRefacter Endpoints into EndpointMap
Mark Chappell [Tue, 3 Nov 2015 13:29:57 +0000 (15:29 +0200)]
Refacter Endpoints into EndpointMap

Because many of the service endpoints URLs use the same patterns for
generating the URLs it makes sense to use the same templates to reduce
the copy and paste.

In the process also adds support for explicitly specifying hostnames
for use in the endpoints.  Note: DNS must be pre-configured.   The
Heat templates do not directly configure DNS.

Change-Id: Ie3270909beca3d63f2d7e4bcb04c559380ddc54d
Co-Authored-By: Juan Antonio Osorio Robles <jaosorior@redhat.com>
8 years agoMerge "Enable glance-api show_image_direct_url for COW"
Jenkins [Wed, 11 Nov 2015 02:32:30 +0000 (02:32 +0000)]
Merge "Enable glance-api show_image_direct_url for COW"

8 years agoMerge "Revert "Manage keystone initialization directly in t-h-t manifests""
Jenkins [Wed, 11 Nov 2015 02:21:17 +0000 (02:21 +0000)]
Merge "Revert "Manage keystone initialization directly in t-h-t manifests""

8 years agoMerge "Align some defaults with with ControllerCount default"
Jenkins [Wed, 11 Nov 2015 02:07:22 +0000 (02:07 +0000)]
Merge "Align some defaults with with ControllerCount default"

8 years agoMerge "Allow customization of the Ceph pool names"
Jenkins [Tue, 10 Nov 2015 23:22:03 +0000 (23:22 +0000)]
Merge "Allow customization of the Ceph pool names"

8 years agoChange the Atomic image name so it's less specific
Ryan Hallisey [Tue, 10 Nov 2015 18:28:30 +0000 (13:28 -0500)]
Change the Atomic image name so it's less specific

The atomic image name in glance was being set to 'fedora-atomic'.
The glance image can be any form of atomic distro so we shouldn't
name this specifically 'fedora-atomic', but instead 'atomic-image'.

Change-Id: Ic539b82b92e3fdd834750e591d8622b7dc85fc6d

8 years agohonor the rabbit user and password provided
Mike Burns [Tue, 10 Nov 2015 18:06:58 +0000 (13:06 -0500)]
honor the rabbit user and password provided

Currently rabbit username and password are defaulted and attempting
to use anything else would result in a failure during deployment.

Change-Id: I8a2e240a19f915309eee45ea3c3368d131af6c1b
Related:  rhbz#1261303

8 years agoSet default start/stop timeout for pcmk services to 95s
Giulio Fidente [Wed, 4 Nov 2015 11:19:50 +0000 (12:19 +0100)]
Set default start/stop timeout for pcmk services to 95s

This change will increase the default start/stop timeout for all
the non-ocf pcmk services to 95s to make sure it allows for at
least 90s to the systemd script to complete the start/stop.

More info at: https://bugzilla.redhat.com/show_bug.cgi?id=1275324

Change-Id: I04f691396a4118b456728a43d71d32ac9a556431

8 years agoPin docker version for atomic at 1.8.2
Jeff Peeler [Mon, 9 Nov 2015 22:40:19 +0000 (17:40 -0500)]
Pin docker version for atomic at 1.8.2

docker-latest now points to 1.9.0, which isn't the version we're
targetting. More importantly, docker-1.9.0 doesn't work since it
complains about /etc not being able to be relabeled. Not sure how to fix
that, but we can save that battle for another time.

Change-Id: I947b7569d9cf40a409253336e51b4dec5ada36f8

8 years agoMerge "Make puppet manifests compliant with Puppet 4.x"
Jenkins [Mon, 9 Nov 2015 18:32:30 +0000 (18:32 +0000)]
Merge "Make puppet manifests compliant with Puppet 4.x"

8 years agoMerge "Add Puppet 4.x lint checks"
Jenkins [Mon, 9 Nov 2015 18:32:21 +0000 (18:32 +0000)]
Merge "Add Puppet 4.x lint checks"

8 years agoAdd support to tag patch port for Cisco Nexus1000v VSM bringup
Shiva Prasad Rao [Thu, 5 Nov 2015 23:40:26 +0000 (15:40 -0800)]
Add support to tag patch port for Cisco Nexus1000v VSM bringup

In some deployments we will need to tag the patch port connecting to
vsm-br in order for traffic to go out. This patch takes passes the vlan
parameter to the puppet.

Change-Id: I18734ae39007985769db9371abe1740e0f2872f7

8 years agoAllow customization of Ceph client user
Giulio Fidente [Thu, 13 Aug 2015 18:10:29 +0000 (20:10 +0200)]
Allow customization of Ceph client user

Previously we enforced the Ceph user used by the OpenStack clients
to be named 'openstack', this change allows for customization
of such a name.

Change-Id: Idef3e1ed4e8e21b645081869b8d6fad2329bdc60

8 years agoAllow customization of the Ceph pool names
Giulio Fidente [Thu, 13 Aug 2015 17:04:17 +0000 (19:04 +0200)]
Allow customization of the Ceph pool names

This is useful in those scenarios were we want to use an external
Ceph deployment with multiple overclouds.

Change-Id: I1749d2a6547f6ce25843709e46a1447e8d42cfff

8 years agoMerge "Add network templates for multiple NIC configuration"
Jenkins [Thu, 5 Nov 2015 16:42:15 +0000 (16:42 +0000)]
Merge "Add network templates for multiple NIC configuration"

8 years agoMake puppet manifests compliant with Puppet 4.x
Gael Chamoulaud [Mon, 21 Sep 2015 13:14:27 +0000 (15:14 +0200)]
Make puppet manifests compliant with Puppet 4.x

- https://docs.puppetlabs.com/puppet/3.8/reference/deprecated_language.html
- Temporary disablement of the pupppet-lint autoload layout check
  failing for ringbuilder.pp. A fix for that will be part of an other patch.

Change-Id: I495825641ab12e7c5789c1405649c356c5bb8051
Signed-off-by: Gael Chamoulaud <gchamoul@redhat.com>
8 years agoAdd Puppet 4.x lint checks
Gael Chamoulaud [Tue, 13 Oct 2015 11:15:15 +0000 (13:15 +0200)]
Add Puppet 4.x lint checks

- This change adds Puppet-lint bits for checking Puppet manifests.

Change-Id: I82869d8f9f6c64ac2a95429f453c3cb76c046974
Signed-off-by: Gael Chamoulaud <gchamoul@redhat.com>
8 years agoAdd network templates for multiple NIC configuration
Dan Sneddon [Wed, 22 Jul 2015 08:34:07 +0000 (01:34 -0700)]
Add network templates for multiple NIC configuration

This change adds a set of network interface configurations for use
with network isolation. The multiple-nics templates includes one
separate NIC per network, and assumes that nic1 is used for the
provisioning network (ctlplane). Also included is an environment
file for including the multiple-nics configuration in a deployment.

This revision changes the ordering of the NICs. By doing that, it
is possible to wire up only a subset of the NICs for the storage
nodes, and it is possilbe to leave the External NIC only configured
on the controllers.

rdo: Updated this commit for static control plane configuration

Co-Authored-By: Rhys Oxenham <roxenham@redhat.com>
Change-Id: Ic878d1ed1a85b5705295d087a743570ca8213504

8 years agoMerge "compute/ceilometer: use internalURL for os endpoint type"
Jenkins [Thu, 5 Nov 2015 14:08:34 +0000 (14:08 +0000)]
Merge "compute/ceilometer: use internalURL for os endpoint type"

8 years agoMerge "Fix dependency for Ceilometer agent when using MySQL"
Jenkins [Thu, 5 Nov 2015 09:53:19 +0000 (09:53 +0000)]
Merge "Fix dependency for Ceilometer agent when using MySQL"

8 years agoMerge "Do not enforce start timeout to 90s for neutron-server"
Jenkins [Wed, 4 Nov 2015 21:29:43 +0000 (21:29 +0000)]
Merge "Do not enforce start timeout to 90s for neutron-server"

8 years agoRevert "Manage keystone initialization directly in t-h-t manifests"
Ben Nemec [Wed, 4 Nov 2015 16:55:34 +0000 (16:55 +0000)]
Revert "Manage keystone initialization directly in t-h-t manifests"

This reverts commit 86d6c1ddc76bad423194e789ffb5474e4e12960e.

This likely has an impact on upgrades, and since we don't
have an upgrade CI job yet I'm concerned that we may have
just broken ourselves.  I would prefer to wait to merge this
until the CI job is in place.

Change-Id: Ib2366cb4b40471a28122f6e9955da9bdb31a53fb

8 years agoMerge "Manage keystone initialization directly in t-h-t manifests"
Jenkins [Wed, 4 Nov 2015 16:43:39 +0000 (16:43 +0000)]
Merge "Manage keystone initialization directly in t-h-t manifests"

8 years agoMerge "Cisco Nexus ML2 MD: Need unique key for server level keys in hieradata"
Jenkins [Wed, 4 Nov 2015 15:30:40 +0000 (15:30 +0000)]
Merge "Cisco Nexus ML2 MD: Need unique key for server level keys in hieradata"

8 years agoDo not enforce start timeout to 90s for neutron-server
Giulio Fidente [Wed, 4 Nov 2015 11:08:23 +0000 (12:08 +0100)]
Do not enforce start timeout to 90s for neutron-server

The same timeout value is set for every pcmk service in [1]

1. https://github.com/openstack/tripleo-heat-templates/blob/master/puppet/manifests/overcloud_controller_pacemaker.pp#L81

Change-Id: I253f6cbc5ccdbf8c46cc537ff9600f201aae540a

8 years agoManage keystone initialization directly in t-h-t manifests
Yanis Guenane [Fri, 2 Oct 2015 10:18:08 +0000 (12:18 +0200)]
Manage keystone initialization directly in t-h-t manifests

This is the second change of a servies of two, it creates the
user, user_role, service and endpoint for:

  * glance
  * nova
  * neutron
  * cinder
  * horizon
  * swift
  * ceilometer
  * heat

Change-Id: I50e792d98a2ba516ff498c58ad402f463c5f7e76

8 years agoCreate keystone roles and admin user from t-h-t manifests
Yanis Guenane [Fri, 2 Oct 2015 10:18:08 +0000 (12:18 +0200)]
Create keystone roles and admin user from t-h-t manifests

Currently keystone initialization happens via os-cloud-config [1].

This commit moves some of that directly into the manifests. This is the
first in a series of two changes to migrate it entirely into t-h-t.

This change focus on implementing what keystone.initialize() was doing
on the tripleoclient [2], creates the admin tenant, user and roles.

It also creates the keystone endpoint itself.

1.  https://github.com/openstack/os-cloud-config/blob/master/os_cloud_config/keystone.py#L128-L158
2.  https://github.com/openstack/python-tripleoclient/blob/master/tripleoclient/v1/overcloud_deploy.py#L462-L527

Change-Id: I98555b707ff9b91c6e218de5dca68106ea05c8ea
Depends-On: Ia4b3244f114dcff746ab89d355ad4933f8fdbddf

8 years agoFix dependency for Ceilometer agent when using MySQL
Juan Antonio Osorio Robles [Mon, 26 Oct 2015 08:37:39 +0000 (10:37 +0200)]
Fix dependency for Ceilometer agent when using MySQL

In HA, when using MySQL as a backend for Ceilometer, the dependencies
set for the Ceilometer central agent depended always on MongoDB; Which
should only be the case if MongoDB is set as a backend.

Change-Id: I6fecfe0564b13e9352313c5a3492505b44d12eaa

8 years agoMerge "Ensure memcached resource uses interleave=true"
Jenkins [Mon, 2 Nov 2015 16:31:47 +0000 (16:31 +0000)]
Merge "Ensure memcached resource uses interleave=true"

8 years agoMerge "Support NFS backend for Glance (via Pacemaker)"
Jenkins [Mon, 2 Nov 2015 12:23:56 +0000 (12:23 +0000)]
Merge "Support NFS backend for Glance (via Pacemaker)"

8 years agocompute/ceilometer: use internalURL for os endpoint type
Emilien Macchi [Fri, 30 Oct 2015 13:39:47 +0000 (22:39 +0900)]
compute/ceilometer: use internalURL for os endpoint type

To let ceilometer access to keystone endpoints, use internalURL instead of
publicURL for security & performances reasons.

Change-Id: I6677f40038d6adf36b4f0375484b6ef2e16869e3

8 years agoMerge "Fix password issue with mysql address for ceilometer"
Jenkins [Thu, 29 Oct 2015 04:42:42 +0000 (04:42 +0000)]
Merge "Fix password issue with mysql address for ceilometer"

8 years agoMerge "Add validation to ping the default gateway(s)"
Jenkins [Fri, 23 Oct 2015 12:07:48 +0000 (12:07 +0000)]
Merge "Add validation to ping the default gateway(s)"

8 years agoFix password issue with mysql address for ceilometer
Juan Antonio Osorio Robles [Fri, 23 Oct 2015 10:51:09 +0000 (13:51 +0300)]
Fix password issue with mysql address for ceilometer

The password was being passed incorrectly to the mysql address used for
ceilometer.

Change-Id: I36d92e199d6d75b58ef8c1b66a2dfbcb7052f948

8 years agoMerge "controller/ceilometer: use internalURL for os endpoint type"
Jenkins [Wed, 21 Oct 2015 21:40:29 +0000 (21:40 +0000)]
Merge "controller/ceilometer: use internalURL for os endpoint type"

8 years agoAdd validation to ping the default gateway(s)
Dan Prince [Thu, 23 Jul 2015 01:51:49 +0000 (21:51 -0400)]
Add validation to ping the default gateway(s)

This patch adds a new function that runs on all nodes so that
all default gateways are ping tested early on during the
installation process (before any puppet is executed)

Note: this change depends on the static control plane network IPs
because some distros allow/create multiple default routes without
this patch (see depends-on below).

Depends-On: Ib267e6dcf2d5ff77f7a82ee20a123965c2d07565

Change-Id: Ida125e79f3e549cae66be9dbd562197afa6400a5

8 years agoEnsure ping returns after 5mins timeout
Giulio Fidente [Wed, 21 Oct 2015 15:26:34 +0000 (17:26 +0200)]
Ensure ping returns after 5mins timeout

This change ensures the validation script returns (with a failure)
if the ping command remains stuck waiting for a response for more
than 5 minutes.

Change-Id: Ib184a01e04fe3c5e3ad13757a355735fff84548d

8 years agoMerge "Sync httpd vhost settings in between pcmk and non-pcmk scenarios"
Jenkins [Wed, 21 Oct 2015 13:42:29 +0000 (13:42 +0000)]
Merge "Sync httpd vhost settings in between pcmk and non-pcmk scenarios"

8 years agoMerge "Allow ctlplane network or ID, and default to "ctlplane""
Jenkins [Tue, 20 Oct 2015 21:38:03 +0000 (21:38 +0000)]
Merge "Allow ctlplane network or ID, and default to "ctlplane""

8 years agoMerge "Update overcloud template description and Count constraints"
Jenkins [Tue, 20 Oct 2015 21:32:32 +0000 (21:32 +0000)]
Merge "Update overcloud template description and Count constraints"

8 years agoAlign some defaults with with ControllerCount default
Steven Hardy [Wed, 7 Oct 2015 16:51:30 +0000 (17:51 +0100)]
Align some defaults with with ControllerCount default

NeutronDhcpAgentsPerNetwork is normally correlated with the
number of controllers, so given that the template
defaults the ControllerCount to 1, 1 seems like a more consistent
value for NeutronDhcpAgentsPerNetwork, and it's consistent with
what we test in CI.

Also NeutronL3HA defaults to False, which is consistent with the default
ControllerCount of 1, but NeutronAllowL3AgentFailover defaults to
True, which I assume makes no sense if there's only 1 controller
and no HA

Change-Id: I4612060979d1e4381800f30e359f58aeef61b861

8 years agoAllow ctlplane network or ID, and default to "ctlplane"
Steven Hardy [Wed, 7 Oct 2015 16:34:25 +0000 (17:34 +0100)]
Allow ctlplane network or ID, and default to "ctlplane"

Currently you always have to pass the ctlplane ID because we're still
using the deprecated network_id property for the neutron port resource.

Since Juno, heat has supported a "network" property, which is used
elsewhere, e.g the nested port stacks, so switch to using it in the
overcloud-without-mergepy template, and flip the default to a more
useful "ctlplane" vs an empty string.

This means the stack create should just work on commonly documented
deployments without requiring any parameter.

Change-Id: Ifcea36d26b795c5e8b80accd8112e23b254127be

8 years agoUpdate overcloud template description and Count constraints
Steven Hardy [Wed, 7 Oct 2015 14:51:57 +0000 (15:51 +0100)]
Update overcloud template description and Count constraints

Currently there's a vague list of services in the description, so instead
describe the roles supported for deployment, and encode the minimum allowed
of one Controller/Compute with zero Storage nodes in the parameter constraints.

Change-Id: Ib4917843f3e4770f0260db72719ed6af0ee8dc13

8 years agoSupport NFS backend for Glance (via Pacemaker)
Jiri Stransky [Tue, 7 Jul 2015 13:16:10 +0000 (15:16 +0200)]
Support NFS backend for Glance (via Pacemaker)

Adds support for NFS backend in Glance by allowing the storage directory
for the 'file' backend to be a mount managed by Pacemaker. Default
behavior is unchanged.

Since the Pacemaker-related parameters are not exposed on top level,
change storage-environment.yaml to use parameter_defaults instead of
parameters.

Depends on a Heat fix for environment file's parameter_defaults to
work well with JSONs and comma delimited lists (see Depends-On).

Change-Id: I6e7e2eaf6919b955650c0b32e1629a4067602c89
Depends-On: I85b13a79dbc97a77e20c0d5df8eaf05b3000815e

8 years agocontroller/ceilometer: use internalURL for os endpoint type
Emilien Macchi [Mon, 21 Sep 2015 19:22:54 +0000 (15:22 -0400)]
controller/ceilometer: use internalURL for os endpoint type

To let ceilometer access to keystone endpoints, use internalURL instead
of publicURL for security & performances reasons.

Ceilometer services (API, agents) will use internalURL endpoint to talk
to other services (keystone, neutron, etc).

Change-Id: I4cb843400f244cd34bbae4bc76371977780c7943

8 years agonova: set catalog_info to InternalURL
Emilien Macchi [Mon, 21 Sep 2015 18:46:24 +0000 (14:46 -0400)]
nova: set catalog_info to InternalURL

In nova.conf, set cinder/catalog_info to 'volumev2:cinderv2:internalURL'
instead of 'volumev2:cinderv2:publicURL'.
So Nova will use internal Cinder endpoint to reach volume API by
using internal network.

Depends-On: Id9e579ca31364d5207d0c1b892d0f7aa7f20f7a8
Change-Id: Ia34f0fe59f662c3ad29ca0178c01ef1570759d57

8 years agoEnsure memcached resource uses interleave=true
Giulio Fidente [Mon, 19 Oct 2015 13:52:44 +0000 (15:52 +0200)]
Ensure memcached resource uses interleave=true

Without interleave Pacemaker will consider memcached as a single
unit on all nodes so it will restart dependencies on every node in
case of failures.

More info at: https://bugzilla.redhat.com/show_bug.cgi?id=1262425

Change-Id: I284cbf33c8453a2abefc862f3d2e0b7f11ff4c3d

8 years agoMerge "Force stop a single node pacemaker on yum update"
Jenkins [Fri, 16 Oct 2015 17:52:32 +0000 (17:52 +0000)]
Merge "Force stop a single node pacemaker on yum update"

8 years agoMerge "Puppet / Compute: allow to run Ephemeral only storage with RBD"
Jenkins [Fri, 16 Oct 2015 14:31:31 +0000 (14:31 +0000)]
Merge "Puppet / Compute: allow to run Ephemeral only storage with RBD"

8 years agoMerge "Allow a deployer to specify HAProxy syslog server address"
Jenkins [Fri, 16 Oct 2015 08:52:44 +0000 (08:52 +0000)]
Merge "Allow a deployer to specify HAProxy syslog server address"

8 years agoSync httpd vhost settings in between pcmk and non-pcmk scenarios
Giulio Fidente [Thu, 15 Oct 2015 16:30:42 +0000 (18:30 +0200)]
Sync httpd vhost settings in between pcmk and non-pcmk scenarios

Moves the vhost_params out of the manifest and into static hiera;
also removes unneeded server_alias parameter as that matched the
vhost servername anyway.

Change-Id: I4b5971b23ef3be9529a59075fa93ccc64af75b9c

8 years agoCisco Nexus ML2 MD: Need unique key for server level keys in hieradata
tiswanso [Tue, 6 Oct 2015 15:33:05 +0000 (11:33 -0400)]
Cisco Nexus ML2 MD: Need unique key for server level keys in hieradata

Due to a limitation in the puppet version used in RHEL7 there is no simple
way to scope a 2nd level hiera hash key with the create_resources + defined
types pattern.  Lack of the .each method support prior to puppet 4.0 is the
problem here.  This template change works around the problem by explicitly
adding the hostname to the hieradata for a server under a nexus switch.

The duplicate server names under different switches is needed for vPC
config scenarios.

Closes-bug: #1506546
Change-Id: I03b866fb440e968c9f86ae93942b687e7165a065

8 years agoSet Django cache backend to Memcached instead of LocMemCache
Giulio Fidente [Thu, 15 Oct 2015 15:07:47 +0000 (17:07 +0200)]
Set Django cache backend to Memcached instead of LocMemCache

Change-Id: Ia2079fc3e350cc677811ebb970cd2b306d6e7040

8 years agoAllow a user to specify a comma separated list of ntp servers
Yanis Guenane [Wed, 7 Oct 2015 15:06:45 +0000 (17:06 +0200)]
Allow a user to specify a comma separated list of ntp servers

This commits aims to allow a user to specify several ntp servers and not
just one.

Example:

openstack overcloud deploy --templates --ntp-server
0.centos.pool.org,1.centos.pool.org

Change-Id: I4925ef1cf1e565d789981e609c88a07b6e9b28de

8 years agoMerge "Set shared secrets, keys and passwords as hidden"
Jenkins [Wed, 14 Oct 2015 09:27:36 +0000 (09:27 +0000)]
Merge "Set shared secrets, keys and passwords as hidden"

8 years agoMerge "Parameterize RabbitMQ FD limit"
Jenkins [Tue, 13 Oct 2015 15:59:42 +0000 (15:59 +0000)]
Merge "Parameterize RabbitMQ FD limit"

8 years agoAdd more components virtual ip mapping into controller.
Yanis Guenane [Tue, 6 Oct 2015 08:42:25 +0000 (10:42 +0200)]
Add more components virtual ip mapping into controller.

Currently only Glance and Heat have their virtual IP passed to the
contrller directly.

This commit adds the same feature for :

 * Ceilometer
 * Cinder
 * Nova
 * Swift

Change-Id: I295d15d7a0aa33175a5530e3b155b0c61983b6ae

8 years agoParameterize RabbitMQ FD limit
Giulio Fidente [Tue, 14 Jul 2015 22:30:12 +0000 (00:30 +0200)]
Parameterize RabbitMQ FD limit

Together with [1] this change permits to parameterize the file
descriptor limit for RabbitMQ for both the Systemd startup script
and the Pacemaker resource agent.

1. https://github.com/puppetlabs/puppetlabs-rabbitmq/commit/20325325b977c508b151ef8036107dcfefdf990b

Closes-Bug: 1474586
Change-Id: I62d31e483641ccb5cf489df81146ecb31d0c423f

8 years agoAllow a deployer to specify HAProxy syslog server address
Yanis Guenane [Fri, 11 Sep 2015 13:25:01 +0000 (15:25 +0200)]
Allow a deployer to specify HAProxy syslog server address

This commit aims to allow a deployer to specify where to send haproxy's
logs. It is backward compatible with what is already in place and send
the logs to the UNIX socket /dev/log

The value specified here will be written in the haproxy.cfg file with
the following behavior

HAProxySyslogAddress: 127.0.0.1 -> log 127.0.0.1 local0
HAProxySyslogAddress: ::1 -> log ::1 local0
HAProxySyslogAddress: /dev/log -> log /dev/log local0 (default)

Change-Id: I46c489a1f424e2219d129f332e64c64019aef850
Depends-On: If7f7c8154e544e5d8a49f79f642e1ad01644a66d

8 years agoPuppet / Compute: allow to run Ephemeral only storage with RBD
Emilien Macchi [Wed, 15 Jul 2015 17:30:18 +0000 (13:30 -0400)]
Puppet / Compute: allow to run Ephemeral only storage with RBD

This patch allows the case where we're not running Ceph to host Persistent
storage (volumes) but just to host Ephemeral storage (VMs).

Before we were only allowing Ephemeral storage on Ceph when also
Persistent storage was using Ceph.

Change-Id: I03b775326e4424de413452f4453d4d88de0083bc

8 years agoSet shared secrets, keys and passwords as hidden
Juan Antonio Osorio Robles [Mon, 12 Oct 2015 10:59:22 +0000 (13:59 +0300)]
Set shared secrets, keys and passwords as hidden

Change-Id: Ieb27729c6b33ffc849d07200ec0d42508214956e
Closes-Bug: #1399793