The old syntax that used % symbols, caused some errors upon a restart
of the nova-api. Replacing the % and parentheses with the double quotes
allows the API to start without errors.
Change-Id: I96c00590390c6caf0a17a43e68ede111684b3110
Closes-Bug: #1683496
The nova allocation ratio settings have been moved from the scheduler to
the main nova class. They are being removed from the scheduler class and
need to be passed to nova differently.
Change-Id: I753c795ef62a505b7b2499b3805acb43d1f79b4d
Closes-Bug: #1673893
Cells v2 setup is mandatory for Ocata. We need to bootstrap them on the
controller nodes and run discover_hosts each deployment.
Change-Id: Id392febd0ffb527f9e8305d347805e9f4f776a7f
Closes-bug: #1670462
Nova placement API is recommended in Newton and is mandatory in Ocata.
Scheduling will not work without it. We configure services according to:
https://docs.openstack.org/developer/nova/placement.html
Change-Id: Ic014ac162f50324c0341cc3013cd093d7125d53e
Closes-bug: #1670354
Earlier metadata_workers option was set to processorcount
fact by default, but with upstream patch [1]
it will be switched to os_workers fact,
so in order to keep the same amount of
nova metadata workers we need to set it directly
in openstack_controller task.
[1] https://review.openstack.org/#/c/383690/
Change-Id: I95bc3d7b5f4c71fc9829b0a58d246ac4ea2fe903
Closes-Bug: #1667626
VMware not supported since Fuel 10. So this commit removes
vmware-related manifests.
Implements: blueprint remove-vmware
Change-Id: If5f0c837fe843f9cfd30648fe6e115267467982b
We misconfigured local cache for services with change
https://review.openstack.org/#/q/Id1034e22d79c3ea6b25575d9bcf8e8750a02365d
Thus, it becomes extremely slow when a controller is down.
With this commit we revert things back to normal with local memcached
for all openstack services leaving keystone memcached shared for tokens
(this was thoroughly tested previously)
Change-Id: I8f6bbf77d27f3d8976985241deb8a948984862f5
Closes-bug: #1657727
Messaging rabbit-* parameters are deprecated in upstream, so
this patch switches all main components to usage of transport_url
Change-Id: I042f68ac1aeb14234898711a25eaebe0836cbb90
Partial-Bug: #1477967
This commit changes override_resources type allowing to
- specify any puppet resource
- pass any number of resources
It also introduces the logic to decide whether override_resources
should create a new resource, if it hasn't been found in the catalog.
The resource overrides as well as additional settings are taken from
Hiera.
This logic essentially allows to create any Puppet resource via
passing the right hash to the Hiera. It replaces/extends the old
'Advanced configuration' feature.
Change-Id: I81099ac41c3e2abdaf6fde3a19161709b7ba4b8c
Blueprint: puppet-resource-override
In order to increase nova performance after failover,
we need to point nova to local memcached instance for keystone tokens
Change-Id: Ibd9ce480201685e4ff92e87490f5beff49272e87
Closes-Bug: #1621541
These patches https://review.openstack.org/#/q/topic:old_authtoken
are removing old puppet authtoken parameters and now we need to use
parameters from keystone::authtoken class in each affected module
Closes-Bug: #1640083
Partial-Bug: #1390099
Change-Id: Ife3ed88237ce735b2a068415b1e57fca93498dcd
* reduce workers_max default limits of a 16 to the $::os_workers, which is limited by [2, 8] interval.
* reduce workers/pool size/overflow limited by F($::processorcount) to F($::os_workers)
Closes-bug: 1629238
Change-Id: I5cc4d70b902eeaa1c9cf42911606eba13dd84aa2
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
* override_resources type extended with additional parameter that allows
to control whether create new resource if it is not in catalog
* override_resources is wrapped with create_resources
* user can put any resource definition in hiera to override its configuration
Change-Id: Ic1f7d856666eb84ced3cef986d8765edf707b481
Blueprint: puppet-resource-override
DocImpact
Exec 'neutron-db-sync' fails without any chance to retry via
additional puppet runs if it was completed successfully on the first
run. Adding retries reduces the likelihood that deployment fails.
Change-Id: I27522de30fc29ef7516e3c9baf36516723ced4a5
Related-bug: #1620268
Signed-off-by: Maksim Malchuk <mmalchuk@mirantis.com>
For the multiple region support, in nova.conf
we need to add parameter "os_region_name" to
specify the region_name for controller's region,
otherwise, nova can not find the right cinder
service to call and volume attach fails.
Change-Id: I7a1497368085380f4a55a66b6f770d3f24cdf7e9
Closes-bug: #1621984
This change adds retries to not fail on 'nova-manage' execution
Closes-Bug: #1619740
Change-Id: I33637b17a18f8809f0a5f3778c57d5066fd56b8d
Signed-off-by: Maksim Malchuk <mmalchuk@mirantis.com>
* Fixes for the openstack tasks and openstack
modules.
* Fixes for fuel tasks.
* Fixes for openstack modules
* Noop tests fixes.
Puppet4 support: common spec
Remove dynamic nova-compute service lookup because the performance
impact of it was too high. Now this value is being statically
chosen by the OS family name.
Puppet4 support: cobbler
* Fix duplicate declarations of systemd services
* upstream dns servers should be an array
Puppet4 support: rsyslog
Fix duplicate rsyslog service
between haproxy and logging tasks
Add tests to ensure that service is present
Puppet4 support: ntp
Fix ntp service duplicate declaration
Puppet4 support: horizon
Fix forgotten include of nova::params
Puppet4 support: apache
Fix broken containment and ordering
Puppet4 support: mysql
Remove duplicate declaration of the access file
Puppet4 support: keystone
user_admin_role should be array
Puppet4 support: openstack controller
Add forgotten include of nova::params
Puppet4 support: server_config
* Check for both undef and empty string value for mtu
* tunel_id_ranges should be array
Puppet4 support: rabbitmq
Add forgotten vhost name variable
Puppet4 support: tools
Rewrite badly writtent spec file variables.
Related-Bug: 1586480
Change-Id: Ia9b474a7076365c0fcad2b55c65e8305477a4647
Don't set auth_version to false if nova_hash['auth_version']
isn't set. It leads to breakage of nova-api service work.
Change-Id: I096d97dcf33f80cc5e288ac511b55964ea48b638
Closes-bug: #1614625
Now we use default values of haertbeat related parameters, and
we need to have an ability to override them using Hiera.
Change-Id: I6d5d26657883aff0f1ff8d3faa1176ff6a8eb220
Closes-Bug: #1576320
Closes-Bug: #1573592
This change reworks how we are managing the m1.micro flavor that is used
by OSTF. As part of this change we are moving the creation out of the
general controller configuration and putting it into the task where we
also manage the image installation used by OSTF. As part of this move,
we are also refactoring the haproxy wait configuration for the nova
backends to their own class with tests so it can be more easily reused.
Change-Id: Ieb0ec8dce76fbcf0979dbca8e50447fe77f0ce90
Closes-Bug: #1602355
This change removes the signing_dir from the nova configuration as it is
for PKI tokens which are no longer supported. Additionally, the
keystone_authtoken/memcached_servers should be configured only for the
api services.
Change-Id: I35b91b4fe65d79a5cd4ed81a584d67a97a1e21d5
Related-Bug: #1600276
This change updates where we calculate the memcache server information
to the globals hiera data and updates all the openstack services to
use the heira memcached_servers key for consistency. Additionally all
api services have been updated to configure the memcached servers for
the keystone_authtoken middleware configuration to improve the
responsiveness of the service.
Change-Id: Id1034e22d79c3ea6b25575d9bcf8e8750a02365d
Closes-Bug: #1597512
Now Mirantis OpenStack uses python dictionary's to store
nova-consoleauth tokens and it fails to connect to VNC on
page reload if we use HA cluster, so we need to use
memcache as cache backend.
Change-Id: I7da25b331f5665ec1b82544f476aebefdfa6113d
Closes-Bug: #1576218
open-iscsi package should be installed always with
cinder packages and it was decided to add it as cinder
dependency. This is already fixed in Debian.
Remove this if totally when UCA packages are updated.
Change-Id: I54fa7a55cb490b38e5f98ef5caa76965948c2f37
Related-bug: #1588799
To ensure that manifests are idempotent we need to pass same
variables from different manifests.
This change will fix issue with ironic-compute task
and openstack-controller task.
Change-Id: Ib486d9d9b8b3fc9b951b75b4b4fc37c484250e6a
Closes-Bug: #1585197
Use defaults for heartbeat_timeout_threshold and
heartbeat_rate parameters for all openstack services
except Neutron: heartbeats don't properly work for it.
Change-Id: I72b2d76f56516e69532549bf6c2c6c468ff47089
Closes-bug: #1584004
* The package contains the OCF script now,
remove it from Fuel's fuel-ha-utils.
* Rework the custom fuel rabbitmq-server OCF wrapper
installation from the package, make it
to invoke the upstream OCF RA from
/usr/lib/ocf/resource.d/rabbitmq/rabbitmq-server-ha
* Address some of TODO/FIXME as well.
Related blueprint upstream-rabbit-ocf
Change-Id: Iaa401685eb5e8dc514809eec426b9e550de609ac
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
metadata_workers option from Nova is not used, so currently a number of
metadata workers on all nodes is equal to number of CPUs.
This patch limits a number of workers using following schemes:
- for controllers: min(max($::processorcount, 2), $workers_max)
- for computes (DVR): min($::processorcount / 8 + 1, $workers_max))
Change-Id: Ifac069347be758ac2c12b890a3b4751db1a12e21
Closes-bug: #1575724
Upstream puppet modules switched to
puppet-oslo usage, but they also changed
defaults. This caused unwanted changes
in behavior. Related bug is mentioned
below. This patch "restores" previous
values of message queue parameters.
Change-Id: I5cf9f697911eeef5093306dbcf66647cebcbf91d
Related-bug: #1573592
Nova scheduler has an ability to detect compute nodes with appropriate huge
pages sizes and schedule instances on them without host aggregates
Change-Id: If093b5a704b745feb9a2164bfc988d1172b8f1c5
Closes-bug: #1569775
* get glance api servers list from one entry point for http(s)
* set the following descending priority for getter:
- hiera glance_api_servers raw list
- get_ssl_property() return
- hiera glance_endpoint ip address
- $management_vip
Change-Id: I80c45d1c6cdabfd8aa239ae61d8c4add5e9e8cbe
Closes-Bug: #1568088
Closes-bug: #1569443
Andreas Jaeger