apex-tripleo-heat-templates.git
8 years agoMerge "Fixes ovs bonding option"
Jenkins [Fri, 6 May 2016 09:55:12 +0000 (09:55 +0000)]
Merge "Fixes ovs bonding option"

8 years agoMerge "Additional parameters for Nuage Neutron plugin integration"
Jenkins [Fri, 6 May 2016 09:43:48 +0000 (09:43 +0000)]
Merge "Additional parameters for Nuage Neutron plugin integration"

8 years agoMerge "Remove calls to ::mysql from the manifests"
Jenkins [Thu, 5 May 2016 15:02:48 +0000 (15:02 +0000)]
Merge "Remove calls to ::mysql from the manifests"

8 years agoLeave start/stop/restart for Keystone and Glance in charge to the role
Giulio Fidente [Thu, 28 Apr 2016 22:00:45 +0000 (17:00 -0500)]
Leave start/stop/restart for Keystone and Glance in charge to the role

Change-Id: I511052dc765788336ffd32dee2118d787fce725d

8 years agoRemove calls to ::mysql from the manifests
Giulio Fidente [Thu, 28 Apr 2016 18:27:13 +0000 (13:27 -0500)]
Remove calls to ::mysql from the manifests

The database will be created by the roles so we don't need to call
::mysql from the manifest.

Change-Id: I2b137cbd6597222a72cf46830f34a93f002c70ef
Depends-On: Id065a9180f1f1a41ab225ec5f755498ec7d9a827

8 years agoMerge "Disable VIPs before stopping cluster during version upgrade"
Jenkins [Wed, 4 May 2016 11:14:00 +0000 (11:14 +0000)]
Merge "Disable VIPs before stopping cluster during version upgrade"

8 years agoMerge "Fix distinguishing between stack-create and stack-update"
Jenkins [Wed, 4 May 2016 10:01:28 +0000 (10:01 +0000)]
Merge "Fix distinguishing between stack-create and stack-update"

8 years agoMerge "Use str_split to compute netmask in _v6 port templates"
Jenkins [Wed, 4 May 2016 09:37:37 +0000 (09:37 +0000)]
Merge "Use str_split to compute netmask in _v6 port templates"

8 years agoMerge "Wire missing RabbitClientPort into Glance API role"
Jenkins [Tue, 3 May 2016 14:32:58 +0000 (14:32 +0000)]
Merge "Wire missing RabbitClientPort into Glance API role"

8 years agoMerge "Updates ControlPlaneSubnetCidr to be a string"
Jenkins [Tue, 3 May 2016 12:12:54 +0000 (12:12 +0000)]
Merge "Updates ControlPlaneSubnetCidr to be a string"

8 years agoDisable VIPs before stopping cluster during version upgrade
Ian Pilcher [Mon, 2 May 2016 21:21:55 +0000 (16:21 -0500)]
Disable VIPs before stopping cluster during version upgrade

If "pcs cluster stop --all" is executed on a controller that
happens to have a VIP on the internal network, pcs may use the
VIP as the source address for communication with another cluster
node.  When pacemaker is stopped this VIP goes away, and pcs never
receives a response from the other node.  This causes pcs to hang
indefinitely; eventually the upgrade times out and fails.

Disabling the VIPs before stopping the cluster avoids this
situation.

Change-Id: I6bc59120211af28456018640033ce3763c373bbb
Closes-Bug: 1577570

8 years agoUse str_split to compute netmask in _v6 port templates
Giulio Fidente [Mon, 18 Apr 2016 16:51:11 +0000 (18:51 +0200)]
Use str_split to compute netmask in _v6 port templates

Change-Id: I282dbc025500b1628d4f08a49b54a2adefd38b5f

8 years agoMerge "Fix the Sahara public endpoint and make it point to the Public VIP"
Jenkins [Fri, 29 Apr 2016 16:36:21 +0000 (16:36 +0000)]
Merge "Fix the Sahara public endpoint and make it point to the Public VIP"

8 years agoMerge "Fix ControllerExtraConfig parameter name"
Jenkins [Fri, 29 Apr 2016 16:31:51 +0000 (16:31 +0000)]
Merge "Fix ControllerExtraConfig parameter name"

8 years agoWire missing RabbitClientPort into Glance API role
Giulio Fidente [Thu, 28 Apr 2016 16:17:57 +0000 (11:17 -0500)]
Wire missing RabbitClientPort into Glance API role

Change-Id: I0ebb5a1e504dd3ffef8ec15c721cf9a9bce6f05b

8 years agoAdditional parameters for Nuage Neutron plugin integration
David Gurtner [Tue, 19 Apr 2016 15:34:38 +0000 (17:34 +0200)]
Additional parameters for Nuage Neutron plugin integration

Add additional parameters, specifically:

* set core plugin to Nuage
* disable service plugins
* disable OVS, l3, metadata and DHCP agent
* rename OVS bridge to alubr0
* include Nuage API extensions

Change-Id: Ia0c201fd3b01cd524e096e6f246d707c6e643944

8 years agoFixes ovs bonding option
Ramon Acedo [Thu, 28 Apr 2016 10:53:13 +0000 (11:53 +0100)]
Fixes ovs bonding option

Default ovs bonding options parameter BondInterfaceOvsOptions was wrongly set to "mode=active-backup" which makes os-net-config fail. This fix changes it to "bond_mode=active-backup".

Change-Id: If3eaac6558c1a15ac09b198f90f05f1a77cf794b
Closes-Bug: #1576137

8 years agoFix the Sahara public endpoint and make it point to the Public VIP
Giulio Fidente [Wed, 27 Apr 2016 22:22:59 +0000 (00:22 +0200)]
Fix the Sahara public endpoint and make it point to the Public VIP

Change-Id: Iff287b9ea46100800e386efb98371be7ab48361f

8 years agoMerge "Run sync in step 3 when $sync_db"
Jenkins [Tue, 26 Apr 2016 19:42:21 +0000 (19:42 +0000)]
Merge "Run sync in step 3 when $sync_db"

8 years agoMerge "Fix the ManagementNetValueSpecs param type"
Jenkins [Tue, 26 Apr 2016 19:31:01 +0000 (19:31 +0000)]
Merge "Fix the ManagementNetValueSpecs param type"

8 years agoMerge "Update .sh references from openstack-keystone to openstack-core"
Jenkins [Tue, 26 Apr 2016 17:02:34 +0000 (17:02 +0000)]
Merge "Update .sh references from openstack-keystone to openstack-core"

8 years agoMerge "Use 'deeper' hiera_hash merge behavior for all roles"
Jenkins [Mon, 25 Apr 2016 19:20:15 +0000 (19:20 +0000)]
Merge "Use 'deeper' hiera_hash merge behavior for all roles"

8 years agoRun sync in step 3 when $sync_db
Giulio Fidente [Fri, 22 Apr 2016 17:19:33 +0000 (19:19 +0200)]
Run sync in step 3 when $sync_db

This will configure the openstack services and run the initial
db sync in step 3 (instead of step 4) for the node for which
$sync_db is true.

Closes-Bug: #1572952
Change-Id: I29012ee0a8b281e4472353ee7d9d44912e8a9b6c

8 years agoFix the ManagementNetValueSpecs param type
Matt Wisch [Fri, 22 Apr 2016 15:09:56 +0000 (11:09 -0400)]
Fix the ManagementNetValueSpecs param type

The ManagementNetValueSpecs param type is currently set to string.
This change sets the param to the correct type of json, allowing the
network value specs to correctly parse.

Example Management Network value spec:

{'provider:physical_network': 'management', 'provider:network_type': 'flat'}

Change-Id: I5b12c7251690368d79a4d00725a9d6e0d5e75af8
Closes-Bug: #1573649

8 years agoMerge "Enable client address in Horizon's logs."
Jenkins [Fri, 22 Apr 2016 08:42:06 +0000 (08:42 +0000)]
Merge "Enable client address in Horizon's logs."

8 years agoFix controller-no-external.yaml in bonded configs.
Dan Sneddon [Thu, 21 Apr 2016 22:45:25 +0000 (15:45 -0700)]
Fix controller-no-external.yaml in bonded configs.

For some reason the controller-no-external.yaml template is configured
for DHCP on the control plane interface. We switched to static control
plane IPs before the controller-no-external.yaml was created (IIRC), so
I'm not sure how that happened. This change brings the
controller-no-external.yaml in line with the rest of the bonded NIC
templates.

Change-Id: I2ac929e241707db72a0beabf9d5cd7fc14b90f76

8 years agoThe Sahara SSL endpoint was announced on the wrong port
Giulio Fidente [Thu, 21 Apr 2016 14:41:17 +0000 (16:41 +0200)]
The Sahara SSL endpoint was announced on the wrong port

Change-Id: I0cab3cdb2189dab3844f2eda52b8697d05ad3447

8 years agoUse 'deeper' hiera_hash merge behavior for all roles
Giulio Fidente [Thu, 18 Feb 2016 12:55:23 +0000 (13:55 +0100)]
Use 'deeper' hiera_hash merge behavior for all roles

This change configures the hiera merge behavior to 'deeper' [1],
which is useful to merge values when the same hiera key is found
in multiple datafiles.

The hiera default 'native' only picks the value from the key with
the highest priority in the hierarchy.

1. https://docs.puppetlabs.com/hiera/1/lookup_types.html#deep-merging-in-hiera--120

Change-Id: I88c764d9af510ffbbad9fcaa4b747655e38255c2

8 years agocomposable glance services
Dan Prince [Thu, 24 Mar 2016 00:14:54 +0000 (20:14 -0400)]
composable glance services

Adds new puppet and puppet pacemaker specific services for
Glance API and Glance Registry.

The Pacemaker templates extend the default glance services and
swap in the pacemaker specific puppet-tripleo profile instead.
In the case of pacemaker glance-registry there is no separate
puppet manifest so only the configuration parameters are maintained
there. (Due to the way the pacemaker glance constraints are written
the pacemaker varients of this service can't be split out...)

Depends-On: Ifc388f7058ccfff2818f531bcbc00c7179874bbc
Change-Id: I00a8c916129af43cda225754eb10370289bb4b41

8 years agoMerge "Increase galera sync timeout in yum_update.sh"
Jenkins [Wed, 20 Apr 2016 08:18:15 +0000 (08:18 +0000)]
Merge "Increase galera sync timeout in yum_update.sh"

8 years agoMerge "Add ping_retry function"
Jenkins [Tue, 19 Apr 2016 15:39:08 +0000 (15:39 +0000)]
Merge "Add ping_retry function"

8 years agoMerge "Make HeatWorkers param affect the engine workers"
Jenkins [Tue, 19 Apr 2016 06:50:52 +0000 (06:50 +0000)]
Merge "Make HeatWorkers param affect the engine workers"

8 years agoMerge "Reload haproxy after injecting certs w/o pcmk too"
Jenkins [Mon, 18 Apr 2016 20:11:06 +0000 (20:11 +0000)]
Merge "Reload haproxy after injecting certs w/o pcmk too"

8 years agoFix ControllerExtraConfig parameter name
Jiri Stransky [Fri, 15 Apr 2016 15:55:30 +0000 (17:55 +0200)]
Fix ControllerExtraConfig parameter name

We've had a typo for a while that a parameter is named
"controllerExtraConfig" with lowercase c, which can be quite confusing
for users because the other similar parameters
(e.g. NovaComputeExtraConfig) consistently start with an upper case
letter.

We'll support both variants from now on, marking the typoed variant as
deprecated.

Change-Id: Ic67a4297e7fa08308889b95ba35389a01f70f5a4

8 years agoEnable client address in Horizon's logs.
Dimitri Savineau [Wed, 23 Mar 2016 15:49:41 +0000 (11:49 -0400)]
Enable client address in Horizon's logs.

Horizon's backends (httpd) see IP address of the haproxy in the logs instead
of the client address.
This patch allows to:
  - Install the remoteip httpd module [1].
  - Use the X-Forwarded-For HTTP header and override the haproxy address.
  - Configure the Horizon's logs with the client address via httpd logformat.

[1] https://httpd.apache.org/docs/2.4/mod/mod_remoteip.html
[2] https://httpd.apache.org/docs/2.4/mod/mod_log_config.html#logformat

Change-Id: Ib2f215913065426848b48f6293f33a75aff3d328
Depends-On: I54f0f5549d64768dacca71539c71a28cc99d9d95

8 years agoMerge "Make sure openstack services are dependent on openstack-core"
Jenkins [Fri, 15 Apr 2016 10:12:57 +0000 (10:12 +0000)]
Merge "Make sure openstack services are dependent on openstack-core"

8 years agoMake HeatWorkers param affect the engine workers
Juan Antonio Osorio Robles [Fri, 15 Apr 2016 08:15:53 +0000 (11:15 +0300)]
Make HeatWorkers param affect the engine workers

With the addition of the num_engine_workers param in puppet-heat, we
can now wire the HeatWorkers parameter to also set that value, as
with the other workers.

Change-Id: I169b648737f797ccd45d12b66705d111e7b95bb7
Depends-On: I53a4333ca16df516294c733f85546c0b0a6a0b88

8 years agoFix Neutron enable_*_agent* Hiera parameters
Emilien Macchi [Thu, 7 Apr 2016 14:57:38 +0000 (10:57 -0400)]
Fix Neutron enable_*_agent* Hiera parameters

This change will fix a logic error when L3 agent was disabled, where
a Pacemaker constraint (neutron-dhcp-agent-to-l3-agent-constraint) were
still looking for l3_agent_service in the Puppet catalog, but could not
because L3 agent was disabled.
It was sending this Puppet error:
Error: Could not find dependency
Pacemaker::Resource::Service[neutron-l3-agent] for
Pacemaker::Constraint::Base[neutron-dhcp-agent-to-l3-agent-constraint]

Change-Id: I0e5d24d844810c58a3205303399d1c20773af3dd

8 years agoMerge "Add GlanceRegistry to the endpoint map"
Jenkins [Thu, 14 Apr 2016 15:18:42 +0000 (15:18 +0000)]
Merge "Add GlanceRegistry to the endpoint map"

8 years agoMerge "Use a different ceph key for admin/client user"
Jenkins [Thu, 14 Apr 2016 13:11:24 +0000 (13:11 +0000)]
Merge "Use a different ceph key for admin/client user"

8 years agoAdd GlanceRegistry to the endpoint map
Dan Prince [Sat, 9 Apr 2016 14:44:17 +0000 (10:44 -0400)]
Add GlanceRegistry to the endpoint map

This patch adds GlanceRegistry to the endpoint map. This
will make accessing Glance registry setings via the endpoint
map possible.

Change-Id: I9186e56cd4746a60e65dc5ac12e6595ac56505f0

8 years agoKeep setting backend_host to 'hostgroup' for rbd backends
Jiri Stransky [Thu, 14 Apr 2016 10:09:34 +0000 (12:09 +0200)]
Keep setting backend_host to 'hostgroup' for rbd backends

There are backwards incompatible patches [1][2] in puppet-cinder which break
upgrade scenarios, but at this point they made it to liberty and mitaka,
so workaround in t-h-t is easier than revert

[1] https://review.openstack.org/#/c/209412/
[2] https://review.openstack.org/#/c/231068/

Change-Id: Ic82258bf0893ebd4e595e5df73ffbc4c6443f9e8
Closes-Bug: #1570265

8 years agoAdd ping_retry function
Dan Prince [Wed, 6 Apr 2016 18:06:14 +0000 (14:06 -0400)]
Add ping_retry function

This patch cleans up the retry logic so that it is all
consolidated in the same function.

Change-Id: I003afe4963e2af8d55f11a4ae4867e563a22134f
Related-bug: #1567011

8 years agoRefactor HAproxy and VIP creation.
Sofer Athlan-Guyot [Wed, 13 Apr 2016 20:42:15 +0000 (22:42 +0200)]
Refactor HAproxy and VIP creation.

This part in overcloud_controller_pacemaker.pp has a lot of duplicate
code to define haproxy and vip creation.  This is an attempt to refactor
this.

Change-Id: Icbd560de08999e48cfb54c6f3c94f8b96cddd6ba
Depends-On: I4cc6711911c1bfa1bc6063979e2b2a7ab5b8d37b

8 years agoMake sure openstack services are dependent on openstack-core
Jiri Stransky [Tue, 12 Apr 2016 11:45:17 +0000 (13:45 +0200)]
Make sure openstack services are dependent on openstack-core

Previously ceilometer-notification, aodh-listener and sahara-engine
didn't have constraints that would anchor them under openstack-core
dummy resource. Such constraints are added now. (sahara-engine starting
after sahara-api, aodh-listener after aodh-evaluator, and
ceilometer-notification after openstack-core.) Openstack-core ->
heat-api constraint has been removed because heat-api depends on
ceilometer-notification, so there's a transitive dependency on
openstack-core already.

Change-Id: Ided7321ebbf2c3556726343b4bb466fd8759b43a
Closes-Bug: #1569444

8 years agoAlways use parameter_defaults in environment files
Jiri Stransky [Fri, 8 Apr 2016 08:59:07 +0000 (10:59 +0200)]
Always use parameter_defaults in environment files

In the environments/ subdirectory of tripleo-heat-templates, we mostly
use parameter_defaults, but some of the environment files still use
parameters. This can lead to confusing behavior with respect to
parameter priority when passing environment files to deploy/update
commands. Users might expect that subsequent environment files take
priority over preceding ones, but that might not be the case if the
preceding environment files use `parameters`, while the subsequent ones
use `parameter_defaults`.

This commit switches all `parameters:` uses in environments/
subdirectory to `parameter_defaults:`.

Change-Id: Ie4c03c7e7f5a5004a0384d35817135f357e9719b
Closes-Bug: #1567837

8 years agoDeploy Gnocchi as a Ceilometer metrics storage backend
Pradeep Kilambi [Sat, 12 Mar 2016 00:24:41 +0000 (19:24 -0500)]
Deploy Gnocchi as a Ceilometer metrics storage backend

* Deploy Gnocchi API.
* Storage backends: swift, rbd and file.
* Indexer backend default to mysql
* Configure Ceilometer to send metrics datas to Gnocchi
* Pacemaker config

Depends-On: Ic8778a3104e0ed0460423e4bf857682220dc5802
Depends-On: I7d2eb9405e0171fc54fa0b616122f69db5f51ce2

Co-Authored-By: Pradeep Kilambi <pkilambi@redhat.com>
Change-Id: Ifde17b1ab8fa2b30544633e455e1c7eb475705aa

8 years agoMerge "composable keystone services"
Jenkins [Mon, 11 Apr 2016 12:51:12 +0000 (12:51 +0000)]
Merge "composable keystone services"

8 years agoFix distinguishing between stack-create and stack-update
Jiri Stransky [Mon, 11 Apr 2016 10:38:00 +0000 (12:38 +0200)]
Fix distinguishing between stack-create and stack-update

Previously we tried to use UpdateIdentifier for two different things:
tell whether to perform package update, and also to tell whether the
top-level stack is being created or updated (which was incorrect and
resulted in bug 1567384, and an attempt to work around that bug resulted
in bug 1567385).

We cannot use Heat's "action" conditionals in some cases, because they
refer to the direct parent stack, which can yield undesirable results
when introducing new nested stacks or temporarily no-opping something
and then adding it back (in both these cases, "action" would be
considered "CREATE", even though the top-level stack is in "UPDATE").

So tripleoclient passes a new parameter StackAction to tell whether the
top-level stack is being created or updated, and we make use of
that. (It seems there's no better way of getting this info from within
the nested Heat stacks.)

Change-Id: Ie14ddbff15e7ed21aaa3fcdacf36e0040f912382
Depends-On: I9dc3b4cd8a6a71df34d8babf0e4c6505041f5311
Closes-Bug: #1567384
Related-Bug: #1567385

8 years agoUpdate .sh references from openstack-keystone to openstack-core
Giulio Fidente [Wed, 23 Mar 2016 16:14:22 +0000 (17:14 +0100)]
Update .sh references from openstack-keystone to openstack-core

The update and upgrade shell scripts were still referencing the
old openstack-keystone service which got removed with
Ie26908ac9bfc0b84b6b65ae3bda711236b03d9d4

Also removes kilo and liberty specific workarounds and config changes.

Change-Id: Icc80904908ee3558930d4639a21812f14b2fd12e

8 years agoMerge "Don't have separate protocols/ports for Keystone v3"
Jenkins [Mon, 11 Apr 2016 11:00:37 +0000 (11:00 +0000)]
Merge "Don't have separate protocols/ports for Keystone v3"

8 years agoMerge "Change /etc/hosts format and add domain"
Jenkins [Mon, 11 Apr 2016 06:29:01 +0000 (06:29 +0000)]
Merge "Change /etc/hosts format and add domain"

8 years agoDon't have separate protocols/ports for Keystone v3
Zane Bitter [Wed, 24 Feb 2016 17:04:59 +0000 (12:04 -0500)]
Don't have separate protocols/ports for Keystone v3

The change in ab068a824ed51e78bf111387223e58e885ec5c84 is described as
temporary, so it would be better if it did not affect the EndpointMap
parameter (which is effectively a public interface, since it may be
overridden in an environment file). No configuration should end up with
different ports/protocols/hosts for Keystone v2 and v3, and somebody
customising them should not have to account for them separately. Nor
should things break when the need to distinguish between v2 and v3
endpoints goes away.

This change removes the KeystoneV3* keys from the EndpointMap input and
uses the Keystone* keys instead, so that any change to the internal
organisation becomes transparent to the user.

Change-Id: If4cdd9232f4dbc9f2af651bbdfe68f09dc26ed2e

8 years agoMerge "Replace extraconfig/tasks/noop.yaml w/ Heat::None"
Jenkins [Mon, 11 Apr 2016 01:14:08 +0000 (01:14 +0000)]
Merge "Replace extraconfig/tasks/noop.yaml w/ Heat::None"

8 years agoMerge "Add EndpointMap as output"
Jenkins [Sat, 9 Apr 2016 15:01:20 +0000 (15:01 +0000)]
Merge "Add EndpointMap as output"

8 years agocomposable keystone services
Dan Prince [Tue, 22 Mar 2016 01:26:53 +0000 (21:26 -0400)]
composable keystone services

Adds new puppet and puppet pacemaker specific services for
Keystone.

The puppet manifests for keystone now live in puppet-tripleo.
Hiera settings are driven by the nested stack heat templates
and used to control puppet-keystone and puppet-tripleo
directly.

The Pacemaker template extends the default keystone service and
swaps in the pacemaker specific puppet-tripleo profile instead.

Change-Id: I8b30438a27e9d5ec4e7d335e0bd1a931a20b03a2
Depends-On: I2faf5a78db802549053ec41678bf83bf28108189

8 years agoMerge "ha/glance-fs: avoid a race condition"
Jenkins [Sat, 9 Apr 2016 08:42:28 +0000 (08:42 +0000)]
Merge "ha/glance-fs: avoid a race condition"

8 years agoReplace extraconfig/tasks/noop.yaml w/ Heat::None
Dan Prince [Fri, 8 Apr 2016 12:06:18 +0000 (08:06 -0400)]
Replace extraconfig/tasks/noop.yaml w/ Heat::None

Removes the old noop nested stack template for extraconfig
tasks and instead uses OS::Heat::None. This should avoid a few
extra resource checks on create and update.

Change-Id: I5a42fc78ece2553e86385236e214aa1e3c91cd85

8 years agoRemove network/noop.yaml, use OS::Heat::None
Dan Prince [Thu, 7 Apr 2016 23:27:57 +0000 (19:27 -0400)]
Remove network/noop.yaml, use OS::Heat::None

Removes the old noop nested stack template for networks and
instead uses OS::Heat::None instead. This should avoid a few
extra resource checks on create and update.

Change-Id: Ia3d7f62dbda2705ffc3d9edcddebcd3ece3cc9d2

8 years agoha/glance-fs: avoid a race condition
Emilien Macchi [Mon, 28 Mar 2016 20:09:20 +0000 (16:09 -0400)]
ha/glance-fs: avoid a race condition

Create the glance-fs Pacemaker resource on one node (pacemaker master)
instead of all nodes, and set verify_on_create to True.

* It will avoid a race condition if Puppet is applied on 2 nodes on the
  same time, so the filesystem is attempted to be created once.
* Verify with psc that the resource has been correctly created.

The full context of the bug is decribed here:
https://bugzilla.redhat.com/show_bug.cgi?id=1319384

Change-Id: I625f0879ae56e814664d1433ae47e27148779f12

8 years agoAdd removal of the /etc/resolv.conf.save file for +bug/1567004
marios [Thu, 7 Apr 2016 13:39:46 +0000 (16:39 +0300)]
Add removal of the /etc/resolv.conf.save file for +bug/1567004

The change at https://review.openstack.org/#/c/302352/ should stop
the if up/down scripts from making changes to resolv.conf as
discussed in that review and the related bug below. However during
upgrades, as we are moving from a version of the ifcfg-vlanXX files
that don't have the PEERDNS=no added by /#/c/302352 the if up
script will restore the /etc/resolv.conf.save to /etc/resolv.conf
and overwrite it. This removes the .save file during the upgrade
init command which gets delivered to all nodes as the first stage
of a major upgrade.

Change-Id: I91dd139f43be4912c20d8661691bee2b662964d4
Related-Bug: 1567004

8 years agoRevert "Ping retry"
Dan Prince [Wed, 6 Apr 2016 18:17:23 +0000 (14:17 -0400)]
Revert "Ping retry"

This reverts commit 570c690bfb118e0cf130b7dbed7992676519ed9b.

This patch broke the ping_gateway_function when using IPv6
network isolation.

Change-Id: I57850a527804f2e753270fd9063d119d41a83b17
Closes-bug: #1567011

8 years agoUse a different ceph key for admin/client user
Dimitri Savineau [Wed, 6 Apr 2016 14:48:12 +0000 (16:48 +0200)]
Use a different ceph key for admin/client user

By default only the admin user key is generated and this key is used
for both admin and openstack user.
Because the mode of the client's key file is 644, any user with a
valid shell on the controller/compute/ceph nodes can made admin
operations on the ceph cluster.
This patch allows to used the random key generated by tripleoclient
for the openstack user.

Change-Id: I771bbee81c0acfe593e92a99ad12d6f1f7f445ef
Closes-bug: #1566927
Depends-On: I404665c09084f0a6cd2d8872940ee90220dc5f69

8 years agoIncrease corosync token timeout
Jiri Stransky [Tue, 5 Apr 2016 17:15:39 +0000 (19:15 +0200)]
Increase corosync token timeout

This might prevent dropping members from corosync cluster on high load
environments. Symptoms of this problem happening can sometimes be found
in corosync log:

dub 05 17:23:45 overcloud-controller-0 corosync[14152]: [MAIN ] Corosync
main process was not scheduled for 3691.8391 ms (threshold is 1320.0000
ms). Consider token timeout increase.

The default in the Puppet manifest is 1 second, which matches the
corosync default, and we override it with hiera to 10 seconds.

Change-Id: I5ea850ada657e5eecafa3e8b28613a0ac48e78f3

8 years agoReload haproxy after injecting certs w/o pcmk too
Ben Nemec [Tue, 5 Apr 2016 17:23:12 +0000 (12:23 -0500)]
Reload haproxy after injecting certs w/o pcmk too

This was accidentally dropped from
Id5ed05b3a20d06af8ae7a3d6f859b03399b0d77d but we should handle the
non-pacemaker case as well.

Change-Id: Ia06746f9c536159cd7b62259e450b3dec331cdb0

8 years agoMerge "Filter for local nodes in check_resource function"
Jenkins [Tue, 5 Apr 2016 16:14:39 +0000 (16:14 +0000)]
Merge "Filter for local nodes in check_resource function"

8 years agoMerge "Ping retry"
Jenkins [Tue, 5 Apr 2016 15:48:26 +0000 (15:48 +0000)]
Merge "Ping retry"

8 years agoMerge "Nova needs the proper volumes to use Cinder"
Jenkins [Tue, 5 Apr 2016 11:21:38 +0000 (11:21 +0000)]
Merge "Nova needs the proper volumes to use Cinder"

8 years agoMerge "Enable galera replication for Mariadb 10.1"
Jenkins [Mon, 4 Apr 2016 12:31:34 +0000 (12:31 +0000)]
Merge "Enable galera replication for Mariadb 10.1"

8 years agoChange /etc/hosts format and add domain
Juan Antonio Osorio Robles [Tue, 15 Mar 2016 10:16:06 +0000 (12:16 +0200)]
Change /etc/hosts format and add domain

Right now, the service-related IPs assosiated with the machine are
registered in the /etc/hosts with different hostnames. This is fine,
except if you need to register that hostname in a third party service
(such as FreeIPA), since the current configuration is not assigning a
domain to those IP addresses. So the current implementation requires
DNS to be properly working, which is not ideal for testing purposes.

Since the current hostnames are not currently being used; it's still
trivial to change this mapping and the format of them. instead of
having entries such as:

    <INTERNAL IP>  <node>-internalapi
    <STORAGE IP>   <node>-storage
    ...

in /etc/hosts; This changes the format to:

    <INTERNAL IP>  <node>.internalapi.<domain>  <node>.internalapi
    <STORAGE IP>   <node>.storage.<domain>  <node>.storage
    ...

So the network (external, internal, storage, etc...) is now
represented as a subdomain. For simplicity, the format without the
domain is still available through an alias.

Change-Id: I6502959a974546e5de757935acea15df6326acda

8 years agoNova needs the proper volumes to use Cinder
Ryan Hallisey [Sat, 19 Mar 2016 12:31:29 +0000 (12:31 +0000)]
Nova needs the proper volumes to use Cinder

Kolla has been using ceph.  For a while, cinder had
iscsi build into it, but it was removed.  In order to
get this to work with containers again, nova-compute and
libvirt containers need /dev and /lib/udev mounted into their
containers.

We also need to copy nova's rootwrap.conf into the nova.
It was missing this config file.

Change-Id: Ie77f56b4576d5393ad3756b0f5ecc3eeff844d1f

8 years agoFilter for local nodes in check_resource function
Raoul Scarazzini [Fri, 1 Apr 2016 13:52:33 +0000 (15:52 +0200)]
Filter for local nodes in check_resource function

While having extra customizations inside a TripleO deployed
Pacemaker environment, say you have instance HA with
pacemaker_remoted or you need to configure an external arbitrator
for something, then the status of the resources for remote nodes
is "Stopped".
This leads to failures while, for example, scaling up.
This fixes the way status is checked, filtering just local nodes.

Co-Authored-By: Giulio Fidente <gfidente@redhat.com>
Change-Id: I8dc25f5d7031c265858afd5a266fda5315ae37a0

8 years agoMerge "Restart haproxy after configuring SSL certs"
Jenkins [Sun, 3 Apr 2016 19:58:31 +0000 (19:58 +0000)]
Merge "Restart haproxy after configuring SSL certs"

8 years agoMerge "Configure ControllerServices via resource chains"
Jenkins [Sun, 3 Apr 2016 15:33:06 +0000 (15:33 +0000)]
Merge "Configure ControllerServices via resource chains"

8 years agoRestart haproxy after configuring SSL certs
Ben Nemec [Thu, 31 Mar 2016 21:42:11 +0000 (16:42 -0500)]
Restart haproxy after configuring SSL certs

If a certificate expires, the user will need to update it.  However,
because we only restart services at the end of a stack-update the
new certificate doesn't take effect until after puppet has run.
This is a problem because puppet makes OpenStack calls, which will
fail if the certificate is expired.  In that case we never get to
the service restart so the stack is wedged until the user manually
restart haproxy.

This patch addresses the problem by reloading haproxy before puppet
runs.  This is done in a pre-puppet script for pacemaker after pacemaker
is maintenance mode because we need to make sure it happens after all of
the certs have been installed on the controllers, but before puppet
runs.

For non-pacemaker, haproxy is simply reloaded.

Change-Id: Id5ed05b3a20d06af8ae7a3d6f859b03399b0d77d

8 years agoMerge "Disable Nova v3 API"
Jenkins [Fri, 1 Apr 2016 15:58:13 +0000 (15:58 +0000)]
Merge "Disable Nova v3 API"

8 years agoDisable Nova v3 API
Jiri Stransky [Thu, 31 Mar 2016 12:31:53 +0000 (14:31 +0200)]
Disable Nova v3 API

Microversions since Nova API v2.1 are aimed to replace the v3 work. The
/v2.1 is backwards compatible with the legacy /v2 endpoint. What we
called in the past /v3 is now something defunct in-tree. The /v2.1 API
is based on the v3 work, but there are many things that differ, in
particular with the backwards-compat thing. We keep the /v2 path in
api-paste.ini for making sure an upgrade doesn't trample operators and
users but if you look in tree, that's redirecting to the v2.1
codepath (just not asking for microversions). In summary, we only need
one endpoint, ie. /v2.1.

Additional information at https://bugzilla.redhat.com/show_bug.cgi?id=1291291

Related-Bug: #1564372
Change-Id: I1654665663bc5a19c201f7d25407910654ac1308
Depends-On: I6d64b8bcd0f79f1f298ddc809e6d92fbc2985c45

8 years agoConfigure ControllerServices via resource chains
Dan Prince [Tue, 15 Mar 2016 14:49:30 +0000 (10:49 -0400)]
Configure ControllerServices via resource chains

This patch wires in a new for Mitaka Heat feature
that allows us to dynamically include a set of nested
stacks representing individual services via a Heat resource chain.
Follow on patches will use this interface to decompose the controller
role into isolated services.

Co-Authored-By: Steve Hardy <shardy@redhat.com>
Depends-On: If510abe260ea7852dfe2d1f7f92b529979483068
Change-Id: I84c97a76159704c2d6c963bc4b26e365764b1366

8 years agoAdd EndpointMap as output
Juan Antonio Osorio Robles [Wed, 30 Mar 2016 15:30:42 +0000 (15:30 +0000)]
Add EndpointMap as output

The endpoint map contains, not only the hosts and protocols that
the resulting services will use, but also the ports. This
information is useful, and the aim of this patch is to make it
available for the tripleoclient to use it.

Change-Id: I4cc5bbf2e7200f78cd90b93659c326a9200278d7

8 years agoUpdates ControlPlaneSubnetCidr to be a string
Rhys Oxenham [Wed, 30 Mar 2016 23:08:17 +0000 (00:08 +0100)]
Updates ControlPlaneSubnetCidr to be a string

Heat is unable to execute the following join when creating the
os-net-config definition (OsNetConfigImpl):

addresses:
-
ip_netmask:
list_join:
- '/'
- - {get_param: ControlPlaneIp}
          - {get_param: ControlPlaneSubnetCidr}

In the default network-environment.yaml environment file the
ControlPlaneSubnetCidr is defaulted as a non-string "24", e.g.

ControlPlaneSubnetCidr: 24

Resulting in the following error, upon deploymnet:

2016-03-30 22:44:17 [overcloud-Controller-wazxxjc6dq22-2-2ictqfdx2nuo]:
  CREATE_FAILED  Resource CREATE failed:
    resources.NetworkConfig:
      Property error:
        resources.OsNetConfigImpl.properties.config:
          Items to join must be strings not 24

This patch simply updates the default value to '24', to match other
templates. This allows a deployment to succeed as expected.

Change-Id: I5f75b02dba5c35a9856c6ceff9cf6c24a0cb56d8

8 years agoRemove hack the pulls latest docker
Ryan Hallisey [Tue, 15 Mar 2016 19:49:10 +0000 (19:49 +0000)]
Remove hack the pulls latest docker

Atomic is set to Docker 1.8.2. We no longer need to pull the
latest Docker to make our template work.

Change-Id: I8ab4e135ed4891763f8ced596116b14101466160
Co-Authored-By: Ian Main <imain@redhat.com>
8 years agoCentos gives /dev/pts/ptmx the wrong perms
Ryan Hallisey [Mon, 21 Mar 2016 23:39:37 +0000 (23:39 +0000)]
Centos gives /dev/pts/ptmx the wrong perms

In order to use cinder, we need to be able to use
/dev/pts/ptmx.  Centos sets this to 000 when on Fedora
it's 666.

Change-Id: I76dc5adc64d2da0d27204ea31175244bc1b94428

8 years agoEnable galera replication for Mariadb 10.1
Damien Ciabrini [Wed, 30 Mar 2016 12:19:39 +0000 (14:19 +0200)]
Enable galera replication for Mariadb 10.1

The generated galera config has to include additional settings for
galera to be active on MariaDB 10.1.

wsrep_on must be explicitely set to ON. On MariaDB 5.5, this was
implicitely set as soon as wsrep_provider was specified.

a valid wsrep_cluster_address must be configured in addition to
wsrep_on, otherwise recovery command mysqld_safe --wsrep-recover
cannot retrieve replication state, and cluster cannot be bootstrapped.

These explicit settings are backward compatible with MariaDB 5.5 since
the two variables exist in both versions of MariaDB.

Change-Id: I4ab4f4eeb8679899f194399ba8695155e9a2f4a5
Closes-Bug: 1563751

8 years agoMerge "compute: restart OVS agent on neutron.conf changes"
Jenkins [Wed, 30 Mar 2016 09:35:19 +0000 (09:35 +0000)]
Merge "compute: restart OVS agent on neutron.conf changes"

8 years agoMerge "change the default satellite tools rpm repo."
Jenkins [Wed, 30 Mar 2016 06:44:56 +0000 (06:44 +0000)]
Merge "change the default satellite tools rpm repo."

8 years agoPing retry
Joe Talerico [Tue, 29 Mar 2016 18:10:20 +0000 (14:10 -0400)]
Ping retry

The single ping method in the validation script is causing for
deployments to fail. When reviewing the network connectivity, we are
finding we actually do have connectivity
( https://gist.github.com/jtaleric/0276a117625e44993be0 ). This patch is
to change the ping count from 1 to 10, to ensure the network is up.

Closes-Bug: 1563521
Change-Id: I9772407554dffa91978a49a16490ef9ed448a054

8 years agocompute: restart OVS agent on neutron.conf changes
Emilien Macchi [Tue, 29 Mar 2016 19:25:20 +0000 (15:25 -0400)]
compute: restart OVS agent on neutron.conf changes

Some options in neutron.conf are used bu OVS agent, like logging &
messaging.
During the upgrade process, you need to restart the agent if these
options change.

We could patch puppet-neutron to add a notify, but the community won't
like it because Neutron OVS agent is not able to restart gracefully
until [1] got merged. Until that, we can fix it in TripleO, where we
suppose Puppet runs happenning during bootstraps and upgrades.

Later, we'll drop this code from here and move it in puppet-neutron.

[1] https://review.openstack.org/#/c/297211

Change-Id: I02b17b66e93331ddfb1a7abd8adff672bc7a32d6
Closes-Bug: #1563437

8 years agoMerge "PLUMgrid Neutron integration"
Jenkins [Tue, 29 Mar 2016 19:19:41 +0000 (19:19 +0000)]
Merge "PLUMgrid Neutron integration"

8 years agochange the default satellite tools rpm repo.
Mike Burns [Mon, 16 Nov 2015 17:19:23 +0000 (12:19 -0500)]
change the default satellite tools rpm repo.

Change-Id: I60ab36b04b8932e4dbee58e21998dc984178b41c
Bugzilla:  https://bugzilla.redhat.com/1275281

8 years agoMerge "Increment step count to include ringbuilder"
Jenkins [Tue, 29 Mar 2016 16:03:05 +0000 (16:03 +0000)]
Merge "Increment step count to include ringbuilder"

8 years agoMerge "Don't restart Pacemaker-managed services from Puppet"
Jenkins [Tue, 29 Mar 2016 15:38:23 +0000 (15:38 +0000)]
Merge "Don't restart Pacemaker-managed services from Puppet"

8 years agoMerge "Add quotes around the cinder_iscsi_ip_address value"
Jenkins [Tue, 29 Mar 2016 15:36:47 +0000 (15:36 +0000)]
Merge "Add quotes around the cinder_iscsi_ip_address value"

8 years agoDon't restart Pacemaker-managed services from Puppet
Jiri Stransky [Tue, 29 Mar 2016 12:11:08 +0000 (14:11 +0200)]
Don't restart Pacemaker-managed services from Puppet

We need to reload/restart services on updates/upgrades to apply any
config changes, but restarting services managed from Pacemaker from
Puppet causes problems.

For now we no-op the restart and rely on the catch-all restart after
Puppet phase.

In the future we should have a service provider for pacemaker resources
that will be using pcs. We still might have to restart services outside
Puppet due to cluster-wide orchestration issues, but we might be able to
do the restarts selectively rather than restart everything.

We also no-op the start/stop commands to be safe, as it also doesn't
make sense for Puppet to try start and stop those services when it
doesn't have knowledge about Pacemaker.

Change-Id: I95e21e10471cd7575f28c095c48150325f1414b3
Closes-Bug: #1562922

8 years agoMerge "Set UpdateIdentifier for upgrade converge, to prevent services down"
Jenkins [Tue, 29 Mar 2016 11:41:19 +0000 (11:41 +0000)]
Merge "Set UpdateIdentifier for upgrade converge, to prevent services down"

8 years agoMerge "Remove password default for AodhPassword"
Jenkins [Mon, 28 Mar 2016 21:32:15 +0000 (21:32 +0000)]
Merge "Remove password default for AodhPassword"

8 years agoIncrement step count to include ringbuilder
Dan Prince [Tue, 22 Mar 2016 20:22:45 +0000 (16:22 -0400)]
Increment step count to include ringbuilder

This patch wires in ringbuilder.pp so that it is always
asserted like the other manifests and it fixes the misaligned
step sequencing in calling our overcloud controller manifests.

Previously it was called as a separate software deployment outside of
the hiera step sequence. This made things confusing in
controller-post.yaml since the deployment names didn't align
with the step hiera variables after step 3. Now that we call it
just like the other modules it should make gradually moving this
code to puppet-tripleo more straightforward as well.

Change-Id: Ibd4f51f65da475bb20a6b08d7bda673f330a5464

8 years agoSet UpdateIdentifier for upgrade converge, to prevent services down
Mathieu Bultel [Thu, 24 Mar 2016 14:06:42 +0000 (15:06 +0100)]
Set UpdateIdentifier for upgrade converge, to prevent services down

We'd like to let the post puppet pacemaker controller services
restart to happen for the convergence step so set the
UpdateIdentifier. However also set the PackageUpdate to noop so the
yum_update.sh doesn't happen.

Since a full haproxy restart is expected, we no longer need the
systemctl reload added at Iae3bad745ecdf952a7a0314fe1375d07eb47c454
so remove that too.

Some more context at
https://bugzilla.redhat.com/show_bug.cgi?id=1321036

Co-Authored-By: marios <marios@redhat.com>
Change-Id: I31c2d97d68c97b435f63863fae2c89f18f99681d

8 years agoRemove password default for AodhPassword
Pradeep Kilambi [Thu, 24 Mar 2016 18:22:27 +0000 (14:22 -0400)]
Remove password default for AodhPassword

Change-Id: Ibf37bfd6150d212fadcc4d2e2e2d0a89cdd76c91

8 years agoMerge "Fix satellite registration for http or https"
Jenkins [Thu, 24 Mar 2016 18:06:00 +0000 (18:06 +0000)]
Merge "Fix satellite registration for http or https"

8 years agoMerge "Add systemctl reload haproxy to the pacemaker_resource_restart.sh"
Jenkins [Thu, 24 Mar 2016 18:00:09 +0000 (18:00 +0000)]
Merge "Add systemctl reload haproxy to the pacemaker_resource_restart.sh"