apex-tripleo-heat-templates.git
8 years agoMerge "Make sure major upgrade script fails."
Jenkins [Thu, 22 Sep 2016 09:52:37 +0000 (09:52 +0000)]
Merge "Make sure major upgrade script fails."

8 years agoMerge "Provide for RAM-constrained environments"
Jenkins [Wed, 21 Sep 2016 23:10:12 +0000 (23:10 +0000)]
Merge "Provide for RAM-constrained environments"

8 years agoMerge "Glance worker count fix"
Jenkins [Wed, 21 Sep 2016 21:21:11 +0000 (21:21 +0000)]
Merge "Glance worker count fix"

8 years agoMerge "Define step input as a Number type"
Jenkins [Wed, 21 Sep 2016 21:20:48 +0000 (21:20 +0000)]
Merge "Define step input as a Number type"

8 years agoMerge "Update capabilities-map.yaml"
Jenkins [Wed, 21 Sep 2016 21:00:21 +0000 (21:00 +0000)]
Merge "Update capabilities-map.yaml"

8 years agoMerge "Set Neutron's metadata_ip to the nova metadata VIP"
Jenkins [Wed, 21 Sep 2016 16:00:39 +0000 (16:00 +0000)]
Merge "Set Neutron's metadata_ip to the nova metadata VIP"

8 years agoDefine step input as a Number type
Steven Hardy [Wed, 21 Sep 2016 13:53:27 +0000 (14:53 +0100)]
Define step input as a Number type

Currently we pass numbers in (hard-coded in post.j2.yaml) but the
SoftwareConfig schema defaults to String.  If puppet requires an
integer number, setting this type may help preserve the type for
the hook.

Change-Id: Ie9227d7adb58ea3c791aa459a1ab5b17ad935919

8 years agoGlance worker count fix
Joe Talerico [Tue, 2 Aug 2016 18:28:55 +0000 (14:28 -0400)]
Glance worker count fix

This patch changes the default value and type of the Glance worker
configuration to allow it to be unset and allow a system dependent
default to be used (e.g. processorcount or some derivative value). The
previous default of 0 would result in a single self contained process,
which while suitable for debugging and testing is not appropriate for
production deployments.

Partial-Bug: #1626126
Change-Id: I58a6a72a581e7083e1dc4e5ca568fdd3fdd6cdf1

8 years agoProvide for RAM-constrained environments
Jiri Stransky [Wed, 21 Sep 2016 11:53:19 +0000 (13:53 +0200)]
Provide for RAM-constrained environments

We hit problems in environments which don't have a lot of RAM (e.g. dev
envs, could be also CI) that Apache ate too much memory due to
too many worker processes being spawned.

This commit allows customizing the Apache MaxRequestWorkers and
ServerLimit directives via Heat parameters. The default stays 256 as
that's the default in the Puppet module, to be suited for production
environments with powerful machines. Also low-memory-usage.yaml
environment file is added, which can be used to make dev/test/CI
overclouds less memory hungry, where the limits are now set to 32.

Change-Id: Ibcf1d9c3326df8bb5b380066166c4ae3c4bf8d96
Co-Authored-By: Carlos Camacho <ccamacho@redhat.com>
Closes-Bug: #1619205

8 years agoMake defaults from roles_data.yaml more robust
Steven Hardy [Wed, 21 Sep 2016 10:10:47 +0000 (11:10 +0100)]
Make defaults from roles_data.yaml more robust

The previous logic left out the default Count completely when it was
zero, which breaks nested validation and it's likely similar problems
would exist with the other optional defaults, so rework it so the
defaulting happens in the jinja2 logic, and document the interfaces
better in roles_data.yaml

Change-Id: I7f2eb4a3a0b43c5d2cd0d001ed3c73f783c95c74
Closes-Bug: #1625760

8 years agoMerge "Enable L3 HA when multiple controllers and no DVR"
Jenkins [Wed, 21 Sep 2016 10:00:42 +0000 (10:00 +0000)]
Merge "Enable L3 HA when multiple controllers and no DVR"

8 years agoMake sure major upgrade script fails.
Sofer Athlan-Guyot [Wed, 7 Sep 2016 09:25:41 +0000 (11:25 +0200)]
Make sure major upgrade script fails.

Running upgrade-non-controller.sh against compute and object storage did
not fail if the /root/tripleo_upgrade_node.sh failed.

This make it harder to detect error in CI system for instance.

Change-Id: I12b7d640547d3b8ec1f70104d159d6052b7638ff
Closes-Bug: 1620973

8 years agoMerge "RabbitMQ threads should be configured dynamically"
Jenkins [Tue, 20 Sep 2016 22:40:58 +0000 (22:40 +0000)]
Merge "RabbitMQ threads should be configured dynamically"

8 years agoSet Neutron's metadata_ip to the nova metadata VIP
Brent Eagles [Tue, 20 Sep 2016 20:51:40 +0000 (18:21 -0230)]
Set Neutron's metadata_ip to the nova metadata VIP

The neutron metadata agent's metadata_ip field is meant to refer to the
nova metadata service, not the local address on the NeutronApiNetwork.

Change-Id: Ibb25a80ea3e66ab3f5cf63c197460d495939778d
Closes-Bug: #1625504

8 years agoAdd nova-metadata template
Juan Antonio Osorio Robles [Tue, 20 Sep 2016 10:25:53 +0000 (13:25 +0300)]
Add nova-metadata template

This is needed because currently we're not generating
nova_metadata_vip or nova_metadata_nodes_ip, and a service profile is
required for that. Unfortunately, currently puppet-nova only deploys
osapi and metadata through the same manifest, so this profile doesn't
really inject any puppet code. We can make this more elegant later.

Change-Id: Id7112111f16d0c749a6203b90e29e6d9f1e4d57e
Closes-Bug: #1625543

8 years agoRabbitMQ threads should be configured dynamically
Michele Baldessari [Tue, 20 Sep 2016 08:11:54 +0000 (10:11 +0200)]
RabbitMQ threads should be configured dynamically

Currently in puppet/services/rabbitmq.yaml we hardcode the thread pool
size to 30 (via the +A30 snippet):
rabbitmq_environment:
    RABBITMQ_SERVER_ERL_ARGS: '"+K true +A30 +P 1048576 -kernel inet_default_connect_options [{nodelay,true},{raw,6,18,<<5000:64/native>>}] -kernel inet_default_listen_options [{raw,6,18,<<5000:64/native>>}]"'

Upstream rabbit has gained the ability to dynamically configure the
number of threads since 3.6.2 via the following commit:
https://github.com/rabbitmq/rabbitmq-server/commit/41ce5ad808863944cd6d62ce7f7e2271f1010582

Given that the default was hardcoded in rabbit from at least 3.4.0 up
until 3.6.2 (see LP bug associated to this commit), we can actually
remove this hardcoded value as it overrides a sane default.

Before the change:
/usr/lib64/erlang/erts-7.3.1/bin/beam.smp -W w -A 64 -K true -A30 -P 1048576 ...

After the change:
/usr/lib64/erlang/erts-7.3.1/bin/beam.smp -W w -A 64 -K true -P 1048576 ...

So effectively with this change we will have the following:
- With older rabbitmq versions we keep the +A30 default
- With rabbitmq versions >= 3.6.2 the thread number is dynamically
  computed to nr_cpus * 16

Change-Id: I8d30c7d141c29fcc439d40fc767498520be7966e
Closes-Bug: #1625486

8 years agoEnable L3 HA when multiple controllers and no DVR
Brent Eagles [Fri, 16 Sep 2016 20:31:00 +0000 (18:01 -0230)]
Enable L3 HA when multiple controllers and no DVR

This patch conditionally enables Neutron L3 HA if there are multiple
controllers but DVR has not been enabled. If the conditions are false,
the value of NeutronL3HA is used.

Change-Id: If1ebeaf417c0da99d833450e394b71cabff2c800
Closes-Bug: #1623155

8 years agoMerge "Add a function to upgrade from full HA to NG HA"
Jenkins [Mon, 19 Sep 2016 17:23:08 +0000 (17:23 +0000)]
Merge "Add a function to upgrade from full HA to NG HA"

8 years agoMerge "Set VNC URL parameters for nova-compute"
Jenkins [Mon, 19 Sep 2016 15:57:19 +0000 (15:57 +0000)]
Merge "Set VNC URL parameters for nova-compute"

8 years agoAdd a function to upgrade from full HA to NG HA
Michele Baldessari [Fri, 26 Aug 2016 14:46:44 +0000 (16:46 +0200)]
Add a function to upgrade from full HA to NG HA

This is the initial work to have a function that migrates a full HA
architecture as deployed in Mitaka to the HA architecture as deployed in
Newton where only a few resources are managed by pacemaker.

The sequence is the following:
1) We remove the desired services from pacemaker's control. The services
   at this point are still running normally via the systemd service as
   invoked by pacemaker
2) We do a "systemctl stop <service>" on all controllers for all the
   services that were removed from pacemaker's control. We do this to make
   sure that during the yum upgrade, the %post sections that call
   "systemctl try-restart" do not take ages, because at this point during
   the upgrade rabbit is down. The only exceptions are "openstack-core"
   and "delay" which are dummy pacemaker resources that do not exist on
   the system
3) We do a "systemctl start <service>" on all nodes for all the services
   mentioned above.

We should probably merge this patch only when newton has branched as it
is very specific to the M/N upgrade.

Closes-Bug: 1617520
Change-Id: I4c409ce58c1a57b6e0decc3cf168b62698b32e39

8 years agoUse osd_pool_default_* puppet parameters when creating the pools
Giulio Fidente [Wed, 14 Sep 2016 16:15:55 +0000 (18:15 +0200)]
Use osd_pool_default_* puppet parameters when creating the pools

While it is possible to override the pg_num, pgp_num and size for
each pool, the defaults are hardcoded. This patch uses as default
the values given via ceph::profile::params::osd_pool_default_*
parameters, if any.

Closes-Bug: 1623590
Change-Id: Iecde772e7f72fd9abedb54cff4b8f2605df8fedd

8 years agoMerge "M/N upgrade sahara-api fails to restart."
Jenkins [Sat, 17 Sep 2016 19:57:02 +0000 (19:57 +0000)]
Merge "M/N upgrade sahara-api fails to restart."

8 years agoMerge "Add fluentd client service"
Jenkins [Sat, 17 Sep 2016 18:50:51 +0000 (18:50 +0000)]
Merge "Add fluentd client service"

8 years agoMerge "Move rabbit's clustering port away from the ephemeral port range"
Jenkins [Sat, 17 Sep 2016 17:38:36 +0000 (17:38 +0000)]
Merge "Move rabbit's clustering port away from the ephemeral port range"

8 years agoM/N upgrade sahara-api fails to restart.
Sofer Athlan-Guyot [Fri, 19 Aug 2016 17:16:33 +0000 (19:16 +0200)]
M/N upgrade sahara-api fails to restart.

Change-Id: I7a041dab8b1b1edc9c80248e1eef3ce7ab272292
Closes-Bug: 1615056

8 years agoMerge "Rework the pacemaker_common_functions for M..N upgrades"
Jenkins [Sat, 17 Sep 2016 17:28:53 +0000 (17:28 +0000)]
Merge "Rework the pacemaker_common_functions for M..N upgrades"

8 years agoSet VNC URL parameters for nova-compute
Juan Antonio Osorio Robles [Sat, 17 Sep 2016 07:34:48 +0000 (10:34 +0300)]
Set VNC URL parameters for nova-compute

These are needed so the computes can advertize the VNC URL correctly.

Change-Id: Ic3eba9fe929ce396b584249eb84415de09ab1b62
Closes-Bug: #1623607

8 years agoMerge "Add mongo config settings in collector service templates"
Jenkins [Sat, 17 Sep 2016 09:33:38 +0000 (09:33 +0000)]
Merge "Add mongo config settings in collector service templates"

8 years agoRework the pacemaker_common_functions for M..N upgrades
marios [Wed, 25 May 2016 08:56:02 +0000 (11:56 +0300)]
Rework the pacemaker_common_functions for M..N upgrades

For N we cannot assume services are managed by pacemaker.
This adds functions to check if a service is systemd or
pcmk managed and start/stops it accordingly. For pcmk,
only stop/disable on bootstrap node for example, whereas
systemd should stop/start on all controllers.

There is also an equivalent change to the check_resource
which has been reworked to allow both pcmk and systemd.

Implements: blueprint overcloud-upgrades-workflow-mitaka-to-newton
Change-Id: Ic8252736781dc906b3aef8fc756eb8b2f3bb1f02

8 years agoMerge "Add NetApp Manila driver integration and tidy up generic"
Jenkins [Sat, 17 Sep 2016 02:53:34 +0000 (02:53 +0000)]
Merge "Add NetApp Manila driver integration and tidy up generic"

8 years agoMerge "Convert AllNodesExtraConfig to support composable roles"
Jenkins [Sat, 17 Sep 2016 02:53:28 +0000 (02:53 +0000)]
Merge "Convert AllNodesExtraConfig to support composable roles"

8 years agoAdd fluentd client service
Lars Kellogg-Stedman [Tue, 9 Aug 2016 20:20:18 +0000 (16:20 -0400)]
Add fluentd client service

This implements support for installing fluentd agents as a composable
service on the overcloud.

Depends-On: I2e1abe4d8c8359e56ff626255ee50c9cacca1940

Implements: tripleo-opstools-centralized-logging
Change-Id: I23b0e23881b742158fcfb6b8c145a3211d45086e

8 years agoMerge "Expose parameter to enable combination alarms"
Jenkins [Fri, 16 Sep 2016 21:09:15 +0000 (21:09 +0000)]
Merge "Expose parameter to enable combination alarms"

8 years agoMerge "Refactor upgrade checks."
Jenkins [Fri, 16 Sep 2016 20:11:47 +0000 (20:11 +0000)]
Merge "Refactor upgrade checks."

8 years agoMerge "Add CephRgw to roles_data.yaml"
Jenkins [Fri, 16 Sep 2016 19:48:23 +0000 (19:48 +0000)]
Merge "Add CephRgw to roles_data.yaml"

8 years agoMerge "Convert UpdateWorkflow to support composable roles"
Jenkins [Fri, 16 Sep 2016 19:31:08 +0000 (19:31 +0000)]
Merge "Convert UpdateWorkflow to support composable roles"

8 years agoMerge "Fix use of batch_create in CephMon major upgrade template"
Jenkins [Fri, 16 Sep 2016 19:08:37 +0000 (19:08 +0000)]
Merge "Fix use of batch_create in CephMon major upgrade template"

8 years agoMerge "Add hyperconverged-ceph environment to include CephOSD on computes"
Jenkins [Fri, 16 Sep 2016 19:08:30 +0000 (19:08 +0000)]
Merge "Add hyperconverged-ceph environment to include CephOSD on computes"

8 years agoMerge "Fix _from_pool_v6.yaml str_split"
Jenkins [Fri, 16 Sep 2016 18:29:32 +0000 (18:29 +0000)]
Merge "Fix _from_pool_v6.yaml str_split"

8 years agoMove rabbit's clustering port away from the ephemeral port range
Michele Baldessari [Fri, 22 Jul 2016 06:40:25 +0000 (08:40 +0200)]
Move rabbit's clustering port away from the ephemeral port range

Currently RabbitMQ cluster uses a predefined port 35672 for clustering.
This port belongs to so-called ephemeral ports range.

Ephemeral ports are the ports kernel assings to application if it
doesn't specify which port to open. So there is a small chance that this
application being started before RabbitMQ itself could grab this port.
While rather unlikely we did see this happen.

Selinux change should already be in place. On my Centos 7 we have:
rabbitmq_port_t                tcp      25672
corenet_tcp_bind_rabbitmq_port(rabbitmq_t)
corenet_tcp_connect_rabbitmq_port(rabbitmq_t)

First noted via:
https://bugzilla.redhat.com/show_bug.cgi?id=1357522

Closes-Bug: #1623818

Depends-On: I0bcd0d063a7a766483426fdd5ea81cbe1dfaa348
Change-Id: I995bd96c2a17614e954ea5bbae4d58998ef420dc

8 years agoAdd mongo config settings in collector service templates
Pradeep Kilambi [Wed, 14 Sep 2016 20:48:01 +0000 (16:48 -0400)]
Add mongo config settings in collector service templates

In scenario where mongo and collector are on separate nodes like as
indicated in the bug, collector should be able to access mongo replset
and other hiera data.

Closes-bug: #1620468

Depends-On: I0bcd0d063a7a766483426fdd5ea81cbe1dfaa348

Change-Id: Iadf4c78fb03da183d19e93c30f78817a3cfed425

8 years agoMerge "Convert deploy steps to jinja2 loop"
Jenkins [Fri, 16 Sep 2016 15:40:49 +0000 (15:40 +0000)]
Merge "Convert deploy steps to jinja2 loop"

8 years agoFix _from_pool_v6.yaml str_split
Giulio Fidente [Fri, 16 Sep 2016 14:17:22 +0000 (16:17 +0200)]
Fix _from_pool_v6.yaml str_split

Previously [1] we updated from_pool_v6 to use str_split but mistakenly
copy/pasting lines referencing an attribute which isn't created in
these templates.

1. I282dbc025500b1628d4f08a49b54a2adefd38b5f

Closes-Bug: 1624412
Change-Id: I409ff5b36eab2a791db4d352dea5b68096c2dc21

8 years agoMerge "Fixes the Ceph upgrade scripts"
Jenkins [Fri, 16 Sep 2016 14:07:28 +0000 (14:07 +0000)]
Merge "Fixes the Ceph upgrade scripts"

8 years agoMerge "Set client protocol for glance registry client"
Jenkins [Fri, 16 Sep 2016 13:23:00 +0000 (13:23 +0000)]
Merge "Set client protocol for glance registry client"

8 years agoFix use of batch_create in CephMon major upgrade template
Mathieu Bultel [Wed, 14 Sep 2016 13:41:18 +0000 (15:41 +0200)]
Fix use of batch_create in CephMon major upgrade template

The batch_create and rolling_update keys were incorrectly defined
as properties of the resource instead of update policies.

Change-Id: I19261adc78e4cdc3616f16221e85490a6b48d47b
Closes-Bug: 1623506

8 years agoAdd CephRgw to roles_data.yaml
Giulio Fidente [Thu, 15 Sep 2016 10:46:48 +0000 (12:46 +0200)]
Add CephRgw to roles_data.yaml

CephRgw defaults to None in the registry, seems like we missed it
in roles_data after a rebase.

Change-Id: I4ce8b160edfb193f5f6226f8295861e6625ef37b

8 years agoFixes the Ceph upgrade scripts
Giulio Fidente [Thu, 15 Sep 2016 13:31:39 +0000 (15:31 +0200)]
Fixes the Ceph upgrade scripts

The Ceph upgrade scripts was failing on the following:

1. a syntax error in an if condition
2. an attempt to read a possibly unbound variable
3. an attempt to chown a directory which might not exist

this change aims at fixing all of the above.

Closes-Bug: 1623942
Change-Id: I9e9d63d4ab7626893aaf2a25dccfcafbb97ccbdf

8 years agoMerge "Unset Keystone public_endpoint"
Jenkins [Fri, 16 Sep 2016 09:48:47 +0000 (09:48 +0000)]
Merge "Unset Keystone public_endpoint"

8 years agoMerge "Populate vnc_api_lib.ini on compute nodes with OpenContrail"
Jenkins [Fri, 16 Sep 2016 08:58:05 +0000 (08:58 +0000)]
Merge "Populate vnc_api_lib.ini on compute nodes with OpenContrail"

8 years agoConvert AllNodesExtraConfig to support composable roles
Steven Hardy [Thu, 8 Sep 2016 11:14:09 +0000 (12:14 +0100)]
Convert AllNodesExtraConfig to support composable roles

This adjusts the interface to OS::TripleO::AllNodesExtraConfig so
it supports custom/composable/optional roles.

Note this does break backwards compatibility, and I can't see any way
to avoid that.  I've converted the in-tree templates, and we'll have
to document carefully and or provide a script (or automated conversion
via mistral perhaps?) to allow folks to easily adjust any out of tree
templates to the new format.

Basically you just have to:

1. Remove all the *_servers parameters, replace with one "servers"
   json parameter

2. Replace references to e.g "controller_servers" with "servers, Controller"
   which does a path-based lookup into the json map provided by overcloud.yaml

Change-Id: I5eebf853646b2f6300d6b542fcd4f43e82d3b413
Partially-Implements: blueprint custom-roles

8 years agoConvert UpdateWorkflow to support composable roles
Steven Hardy [Thu, 8 Sep 2016 10:51:20 +0000 (11:51 +0100)]
Convert UpdateWorkflow to support composable roles

We need to remove the hard-coded roles from overcloud.j2.yaml
as now it's valid to e.g remove BlockStorage completely.

The previous behavior for the per-role upgrade scripts is maintained
but we'll need to rework this for newton->ocata upgrades where we
can no longer be sure the servers mapping will contain all roles.

Change-Id: I25e6c84757e3c00fba2aae834cd8206c62e44acf
Partially-Implements: blueprint custom-roles

8 years agoConvert deploy steps to jinja2 loop
Steven Hardy [Wed, 7 Sep 2016 16:05:56 +0000 (17:05 +0100)]
Convert deploy steps to jinja2 loop

Refactor so the post-deploy steps recently moved into
puppet/post.yaml are generated by jinja2 instead of hard-coded

Change-Id: I488e46aaa449c95571bd3d1de9513c3d0730baf3
Partially-Implements: blueprint custom-roles

8 years agoSet client protocol for glance registry client
Juan Antonio Osorio Robles [Wed, 14 Sep 2016 10:50:08 +0000 (13:50 +0300)]
Set client protocol for glance registry client

To communicate to glance registry, glance API has several parameters
that it uses to form the URI. Right now we are defaulting to http,
when we enable TLS everywhere, this will break. So setting the value
from the endpoint map should fix it.

Closes-Bug: #1623477
Change-Id: Id86787cbaa6f87fdcf9c26111c228fd59fbba012

8 years agoExpose parameter to enable combination alarms
Pradeep Kilambi [Wed, 31 Aug 2016 15:26:50 +0000 (11:26 -0400)]
Expose parameter to enable combination alarms

The puppet-tripleo change for the same is merged
I9220b7d020dc8ed45dd6ca83ea9647efd67ea648

Change-Id: Ic5309ada98c78a15aa3a47dd94acb9e68eb25295

8 years agoMerge "Convert allNodesConfig properties to composable jinja2"
Jenkins [Wed, 14 Sep 2016 11:32:58 +0000 (11:32 +0000)]
Merge "Convert allNodesConfig properties to composable jinja2"

8 years agoMerge "Add support for deploying Ceph RGW role"
Jenkins [Wed, 14 Sep 2016 09:09:36 +0000 (09:09 +0000)]
Merge "Add support for deploying Ceph RGW role"

8 years agoConvert allNodesConfig properties to composable jinja2
Steven Hardy [Mon, 5 Sep 2016 22:02:44 +0000 (23:02 +0100)]
Convert allNodesConfig properties to composable jinja2

To support custom roles we need to generate these lists of role
specific data.

Change-Id: Ide97cd57d1c07f7f7ff260ff7a6bbe2b71753bd0
Partially-Implements: blueprint custom-roles

8 years agoMove role ResourceGroups inside the jinja2 loop
Steven Hardy [Mon, 5 Sep 2016 21:57:43 +0000 (22:57 +0100)]
Move role ResourceGroups inside the jinja2 loop

This moves the now nearly identical group resources inside the loop
there's a FIXME related to some deprecated compute parameters we'll
need to work around.

Change-Id: Iddd63c42754867125e65e7721ab9d9f46f4d6afb
Partially-Implements: blueprint custom-roles

8 years agoMerge "Enable proxy header parsing for Manila"
Jenkins [Tue, 13 Sep 2016 13:40:39 +0000 (13:40 +0000)]
Merge "Enable proxy header parsing for Manila"

8 years agoAdd NetApp Manila driver integration and tidy up generic
marios [Thu, 11 Aug 2016 11:30:04 +0000 (14:30 +0300)]
Add NetApp Manila driver integration and tidy up generic

Enables configuring a NetApp backend for the Manila service
This was created based on the review at
https://review.openstack.org/#/c/188138/

This makes the netapp and generic backends disabled by default
in the services/manila-backend-*.yaml. A backend is then
enabled via backend-specific environment files, which will set
any config parameters and enable that backend.

It is expected that multiple manila backend specific environment
files might be specified simultaneously.

Finally generic and manila config is split into separate
service files rather than using manila-base for all the things.

Co-Authored-By: Ryan Hefner <rhefner@redhat.com>
Co-Authored-By: Ben Swartzlander <ben@swartzlander.org>
Closes-Bug: 1618479
Depends-On: Ic6f8e8d27ca20b9badddea5d16550aa18bff8418
Change-Id: I35fce32d0f6a5cc1c3382c2d0e0d6028928fd943

8 years agoMerge "De-bracket vncproxy_host in compute profile"
Jenkins [Mon, 12 Sep 2016 22:13:19 +0000 (22:13 +0000)]
Merge "De-bracket vncproxy_host in compute profile"

8 years agoMerge "Configure Keystone credentials"
Jenkins [Mon, 12 Sep 2016 21:52:55 +0000 (21:52 +0000)]
Merge "Configure Keystone credentials"

8 years agoMerge "Add trunking plugin to list of default ML2 service plugins"
Jenkins [Mon, 12 Sep 2016 21:05:24 +0000 (21:05 +0000)]
Merge "Add trunking plugin to list of default ML2 service plugins"

8 years agoUnset Keystone public_endpoint
Adam Young [Mon, 12 Sep 2016 16:43:39 +0000 (12:43 -0400)]
Unset Keystone public_endpoint

The keystone public_endpoint value should be deduced from the calling
request and not hardcoded, or it makes network isolation impossible.

Change-Id: Ide6a65aa9393cb84591b0015ec5966cc01ffbcf8
Closes-Bug: 1381961

8 years agoDe-bracket vncproxy_host in compute profile
Ben Nemec [Tue, 30 Aug 2016 20:58:43 +0000 (15:58 -0500)]
De-bracket vncproxy_host in compute profile

This is done in the vncproxy profile, but for some reason is not in
the compute one.  It causes hiera to explode when the brackets are
left, so we need to do the bracket stripping here too.

Also switches both places to just use the host_nobrackets version
of the endpoint instead of stripping them with str_replace.

Change-Id: I7ccd84b575fd652f6412fdb1869c31c79a7bf53b
Closes-Bug: 1618623

8 years agoConfigure Keystone credentials
Emilien Macchi [Tue, 6 Sep 2016 22:24:28 +0000 (18:24 -0400)]
Configure Keystone credentials

Configure Keystone credentials by installing 2 keys with dynamic content
generated by python-tripleoclient.

Note: this is a first iteration of managing Keystone credentials. It has
a few limitations:
- keys are not exported to external storage.
- keys are not rotated automatically.

Change-Id: I45cf8821eadf528dfcdc8d74e6e0484597b0d2c0

8 years agoAdd explicit entry for ctlplane IP in /etc/hosts
Juan Antonio Osorio Robles [Mon, 12 Sep 2016 11:02:07 +0000 (14:02 +0300)]
Add explicit entry for ctlplane IP in /etc/hosts

There was currently no way of getting it and we can't asure that the
primary IP will use it. So it's explicitly needed there.

Change-Id: Idb3ca22ac136691b0bff6f94524d133a4fa10617

8 years agoRefactor upgrade checks.
Sofer Athlan-Guyot [Thu, 25 Aug 2016 09:58:56 +0000 (11:58 +0200)]
Refactor upgrade checks.

We make it clear that recoverable checks happen before starting the
upgrade to be able to run the upgrade after the offending error has been
manually corrected.

Add new check for the pcsd cluster status.

Add new check for galera password file: BZ 1357112

Closes-Bug: 1614907
Change-Id: If736c79121e1ffe0eaeb814bdb73ccbc0b64edcd

8 years agoMerge "Fix service config files having wrong map_merge format"
Jenkins [Mon, 12 Sep 2016 09:39:05 +0000 (09:39 +0000)]
Merge "Fix service config files having wrong map_merge format"

8 years agoMerge "Make Nova VNC Proxy service name match service net map"
Jenkins [Mon, 12 Sep 2016 07:28:20 +0000 (07:28 +0000)]
Merge "Make Nova VNC Proxy service name match service net map"

8 years agoMerge "Add Heat's Cfn and Cloudwatch networks to ServiceNetMap"
Jenkins [Mon, 12 Sep 2016 06:48:24 +0000 (06:48 +0000)]
Merge "Add Heat's Cfn and Cloudwatch networks to ServiceNetMap"

8 years agoEnable proxy header parsing for Manila
Juan Antonio Osorio Robles [Mon, 12 Sep 2016 06:35:57 +0000 (09:35 +0300)]
Enable proxy header parsing for Manila

This is necessary for when HAProxy is terminating TLS for manila,
else we will have keystone discovery errors. This is the same we do
for several other services, as manila uses the same middleware.

Change-Id: Ice78b0abceb6a956bb8c1dc6212ee1b56b62b43f

8 years agoMove per-role NetIpListMap's into jinja template loop
Steven Hardy [Thu, 1 Sep 2016 13:15:21 +0000 (14:15 +0100)]
Move per-role NetIpListMap's into jinja template loop

Change-Id: I8fc855833e8c602e94d0e8b330a713de1c98f901

8 years agoMerge "Adjust MTU to compensate for VLAN tag issue"
Jenkins [Sat, 10 Sep 2016 14:10:26 +0000 (14:10 +0000)]
Merge "Adjust MTU to compensate for VLAN tag issue"

8 years agoAdd support for deploying Ceph RGW role
Keith Schincke [Sun, 6 Mar 2016 14:33:41 +0000 (09:33 -0500)]
Add support for deploying Ceph RGW role

This patch add support for deploying Ceph RGW.

Co-Authored-By: Giulio Fidente <gfidente@redhat.com>
Change-Id: I88c8659a36c2435834e8646c75880b0adc52e964

8 years agoAdd Heat's Cfn and Cloudwatch networks to ServiceNetMap
Juan Antonio Osorio Robles [Thu, 8 Sep 2016 08:35:02 +0000 (11:35 +0300)]
Add Heat's Cfn and Cloudwatch networks to ServiceNetMap

these are needed if we want to be able to generate which nodes the
service and which VIP are they binding to.

Change-Id: I9d40459406f15db3ad9229c75392c4d959d44b3b
Closes-Bug: #1621371

8 years agoMake Nova VNC Proxy service name match service net map
Juan Antonio Osorio Robles [Thu, 8 Sep 2016 08:21:16 +0000 (11:21 +0300)]
Make Nova VNC Proxy service name match service net map

If these names don't match then we cannot set the service's nodes,
VIP and network.

Change-Id: I8f1c0eaf62eee2704a5f2556a553032106db606b
Closes-Bug: #1621368

8 years agoMerge "Move AllNodesValidationDeployments into jinja template loop"
Jenkins [Sat, 10 Sep 2016 07:55:30 +0000 (07:55 +0000)]
Merge "Move AllNodesValidationDeployments into jinja template loop"

8 years agoMerge "Move AllNodesDeployments into jinja template loop"
Jenkins [Sat, 10 Sep 2016 07:55:23 +0000 (07:55 +0000)]
Merge "Move AllNodesDeployments into jinja template loop"

8 years agoMerge "Convert overcloud.yaml to support jinja2 templating"
Jenkins [Sat, 10 Sep 2016 07:34:58 +0000 (07:34 +0000)]
Merge "Convert overcloud.yaml to support jinja2 templating"

8 years agoMerge "Add support for configuring the OVS firewall driver"
Jenkins [Sat, 10 Sep 2016 02:54:07 +0000 (02:54 +0000)]
Merge "Add support for configuring the OVS firewall driver"

8 years agoMerge "add composable services for Contrail"
Jenkins [Fri, 9 Sep 2016 23:19:12 +0000 (23:19 +0000)]
Merge "add composable services for Contrail"

8 years agoAdjust MTU to compensate for VLAN tag issue
Brent Eagles [Fri, 9 Sep 2016 21:07:39 +0000 (18:37 -0230)]
Adjust MTU to compensate for VLAN tag issue

Some network configurations uncovered what appears to be an issue where
a spurious 802.1q header is injected into tunnelled traffic. Adjusting
the default value to accomodate the extra overhead should avoid this
problem.

Partial-Bug: #1621533
Change-Id: I9ebad2d6ad34d90fcb998497873059995cdef276

8 years agoMove AllNodesValidationDeployments into jinja template loop
Steven Hardy [Mon, 4 Jul 2016 17:56:19 +0000 (18:56 +0100)]
Move AllNodesValidationDeployments into jinja template loop

These are identical for all roles, so move them into the per-role
loop

Partially-Implements: blueprint custom-roles
Change-Id: Id85b830a0e225912a3ea8c8b17a11fc424f68bb0

8 years agoMove AllNodesDeployments into jinja template loop
Steven Hardy [Fri, 1 Jul 2016 16:33:07 +0000 (17:33 +0100)]
Move AllNodesDeployments into jinja template loop

These are identical for all roles, so move them into the per-role
loop

Partially-Implements: blueprint custom-roles
Change-Id: I0a9918d5a2e9a73fe3ac68a96bdee02e95799bc1

8 years agoConvert overcloud.yaml to support jinja2 templating
Steven Hardy [Fri, 1 Jul 2016 14:18:32 +0000 (15:18 +0100)]
Convert overcloud.yaml to support jinja2 templating

The first step of generating the Service chain resources via j2,
we'll then incrementally convert other resources to be created
in a similar way.

Partially-Implements: blueprint custom-roles
Depends-On: I81239991f36ed5f6453184bf9cffe930832cb68b
Change-Id: Iafa9b2afddf18a5a9833ec472a552fb256338b38

8 years agoMerge "Generate VIP info for ctlplane VIP, not management"
Jenkins [Fri, 9 Sep 2016 18:53:07 +0000 (18:53 +0000)]
Merge "Generate VIP info for ctlplane VIP, not management"

8 years agoMerge "Remove *ExtraConfig parameters from overcloud.yaml"
Jenkins [Fri, 9 Sep 2016 16:00:51 +0000 (16:00 +0000)]
Merge "Remove *ExtraConfig parameters from overcloud.yaml"

8 years agoMerge "Move role deployment steps into puppet/post.yaml"
Jenkins [Fri, 9 Sep 2016 15:59:59 +0000 (15:59 +0000)]
Merge "Move role deployment steps into puppet/post.yaml"

8 years agoMerge "Add keystone networks for the different endpoints"
Jenkins [Fri, 9 Sep 2016 11:28:14 +0000 (11:28 +0000)]
Merge "Add keystone networks for the different endpoints"

8 years agoRemove *ExtraConfig parameters from overcloud.yaml
Steven Hardy [Mon, 5 Sep 2016 21:21:22 +0000 (22:21 +0100)]
Remove *ExtraConfig parameters from overcloud.yaml

These per-role parameters are already handled inside the role templates

Change-Id: Ie71c8670ea427ea9c7d94680cff143ad9d524b7a
Partially-Implements: blueprint custom-roles

8 years agoMove role deployment steps into puppet/post.yaml
Steven Hardy [Mon, 5 Sep 2016 17:40:35 +0000 (18:40 +0100)]
Move role deployment steps into puppet/post.yaml

To enable steps to be aligned between roles, we need to define
dependencies between the steps, which is only possible if we
move the steps out of distinct nested stacks so we can use
depends_on to serialized the steps for all roles.

Note that we may be able to further refactor later to remove the
per-role -config.yaml nested stacks as well.

Change-Id: Ia2ea559e8eeb64763908f75705e3728ee90b5744
Partially-Implements: blueprint custom-roles

8 years agoMerge "Add global_config_settings to services' output"
Jenkins [Fri, 9 Sep 2016 09:20:21 +0000 (09:20 +0000)]
Merge "Add global_config_settings to services' output"

8 years agoMerge "Add service networks to hieradata"
Jenkins [Fri, 9 Sep 2016 07:46:08 +0000 (07:46 +0000)]
Merge "Add service networks to hieradata"

8 years agoMerge "Add Ceph cluster health validation on upgrade"
Jenkins [Fri, 9 Sep 2016 06:39:11 +0000 (06:39 +0000)]
Merge "Add Ceph cluster health validation on upgrade"

8 years agoGenerate VIP info for ctlplane VIP, not management
Juan Antonio Osorio Robles [Thu, 8 Sep 2016 13:11:53 +0000 (16:11 +0300)]
Generate VIP info for ctlplane VIP, not management

The management network does not have a VIP, so it's been wrong to
generate a cloud name and hieradata for this. Instead, the network
that actually needs a name and a hosts entry is the ctlplane network,
which actually has a VIP and there are services that use it.

bp tls-via-certmonger

Closes-Bug: #1621742
Change-Id: I163b2c7b5684da6dc290636f54eefe3f2b0c3e3f

8 years agoConvert SwiftDevicesAndProxyConfig to composable format
Steven Hardy [Thu, 1 Sep 2016 18:24:05 +0000 (19:24 +0100)]
Convert SwiftDevicesAndProxyConfig to composable format

Make use of the new composable per-service node_ips lists by
adding a ServiceNetMap entry for SwiftStorage, then
pass the data to construct the raw device list into puppet-tripleo
instead of mangling it in t-h-t inside the role templates.

This will allow running swift storage services on nodes other than
the Controller and ObjectStorage roles, and is required to enable
custom roles.

Depends-On: I11deed1df712ecccf85d36a75b3bd2e9d226af36
Change-Id: I1bf5f8a9d34b1a5d64ab8656b386226b54ec1a27
Partially-Implements: blueprint custom-roles

8 years agoMerge "Add base neutron service configuration"
Jenkins [Thu, 8 Sep 2016 17:44:59 +0000 (17:44 +0000)]
Merge "Add base neutron service configuration"

8 years agoPopulate vnc_api_lib.ini on compute nodes with OpenContrail
Jiri Stransky [Thu, 8 Sep 2016 16:23:09 +0000 (18:23 +0200)]
Populate vnc_api_lib.ini on compute nodes with OpenContrail

This is setting sane defaults for vnc_api_lib.ini as requested from the
field. The settings still can be overriden using NovaComputeExtraConfig
if needed.

Change-Id: I6a823c0b34f6ea21aa16939577ac0e1563483557
Closes-Bug: #1620647