The LMA collectors can push logs before the creation of index template
which leads to have the first index with a wrong number_of_replicas and
wrong field indexation settings.
Closes-Bug: #1632590
Change-Id: Id166f6e0aea67b60c5c38cc6a4a69ece4928634e
This patch removes a dependency that causes a cycle in the graph of
tasks in MOS 8. The cycle was <custom>-netconfig -> hosts -> cluster ->
virtual_ips -> <custom>-netconfig.
We cannot remove hosts.pp as we tried previously because doing this also
remove a dependency between globals and hosts. So we create the
following dependency: tools -> <custom>-netconfig -> hosts
in place of the former: tools -> netconfig -> hosts
Change-Id: I9a2ec06e3f61942ad33b9a32d9a6034602a51abd
Partial-Bug: #1616391
This patch removes a dependency that causes a cycle in the graph of
tasks in MOS 8.
The cycle was <custom>-netconfig -> hosts -> cluster -> virtual_ips ->
<custom>-netconfig
Change-Id: I80f88ebec732c2e95bfa5c271ffb9c35ad0a1827
This change makes sure that the virtual router on the management
network is up and running before configuring the network.
Change-Id: I30063d09d1f7f4e8c5b75beb4d901e174a76f7f1
Closes-Bug: #1583994
Without this change, the nodes are configured to use the Fuel node as
their default gateway. Note that the code path is different between MOS
8 and MOS 9 because of this bug fix [1] that hasn't been backported on
MOS 8.
[1] https://bugs.launchpad.net/fuel/+bug/1541309
Change-Id: I80a2fadf8b16b139d07c48a33c52efd33c66935f
Closes-Bug: #1583994
This patch configures HAProxy to terminate SSL connection when the
support for SSL/TLS is enabled in StackLight.
DocImpact Add support for TLS
Change-Id: Icdff278875d7daf928fa036ec1e4905205791546
Implements: blueprint support-secure-communication
Previously the LMA collector had an explicit dependency on the last
Elasticsearch task but this doesn't work when the Elasticsearch/Kibana
plugin isn't deployed in the same environment as the collector plugin.
This change makes sure that the LMA collector plugin's tasks will only
be executed once the Elasticsearch cluster is fully functional by
pinning both plugins to a well-known task (update_hosts in that case).
Partial-Bug: #1573087
Change-Id: I5bbc444eb51da4787b5be5cbe3aa0284ce6e1753
MOS 9 comes with a new orchestration engine (so-called v2 or task-based
deployement) which reduces greatly the time necessary to deploy new
environments.
This change adds version to the deployment tasks to be fully compatible
with the v2 engine.
Change-Id: Iea7528712b65ce1430c1d8bcb8386908f485a757
This patch adds a new manifest to replace the one that is used from
osnailyfacter. We cannot use the manifest provided by fuel-library
because it has been rewritten and it is not compatible with our
deployment because it requires specific VIP that are tied to the
management and public VIP addresses of the controllers.
Partial-Bug: #1554521
Change-Id: I2e70a19bc1699c61c9fa130e2f4a852e545b1060
This change drops the download of the fuel-library Puppet modules into
the plugin package. Instead the Puppet tasks look for Puppet modules
first in the plugin's path and alternatively in the default
/etc/puppet/module path.
Change-Id: Icf49b48462956fbb57c087e133336738cda70d68
This change makes sure that Pacemaker and HAProxy are reconfigured
on the nodes that are already deployed when the environment is scaled
up. We use the 'update_required' stanza to enforce that the nodes
already deployed are updated in the same run than the new nodes. The
'reexecute_on' stanza doesn't guarantee that.
Change-Id: Id39a6246ee96a693c39898b84092710eb8a97536
Closes-Bug: #1536533
When the cluster does not have quorum (less than 3 Elasticsearch nodes) the
'no-quorum policiy' is set to 'ignore', which allows to continue all resource
management.
If the cluster have quorum (more that 2 nodes), the policy is set to 'stop',
which stops all resources in the affected cluster partition if the cluster
loses quorum to avoid data corruption.
http://clusterlabs.org/doc/en-US/Pacemaker/1.0/html/Pacemaker_Explained/s-cluster-options.html
Change-Id: I3b1ad071e4bf31581f245bbd8de863e6266ffb5b
This change upgrades the package version to 4.0.0 to use the reexecute_on
option.
All tasks defined in tasks.yaml are moved in deployment_tasks.yaml
to leverage this new feature.
The deployment order with 'priority' cannot be set but the only
requirement this plugin has is to be deployed before lma_collector,
which is always the case as long as lma_collector is deployed at
post_deployment time.
Change-Id: I08259f1646122aed0674610ddaf7a327d31b9a1a
- Add firewall rules for corosync communication
- Use dedicated cluster.pp manifest to allow the deployment of coexistant
clusters for LMA plugins
- Use primary role property
Implements: blueprint elasticsearch-clustering
Change-Id: Ibf4c1c4e62f214725875869621b40a3ef4c20e53
Configure a VIP and Corosync/Pacemaker cluster. The plugin must override
Hiera data to set explicitly the corosync node list.
Configure Elasticsearch instances with parameters:
* cluster.name
* unicast discovery
Add firewall rule to allow clustering traffic.
Implements: elasticsearch-clustering
Change-Id: I0636e02113bfdacc776beb20c08cc88308486c29
- Use plugin version 3.0.0 and remove compatibility with MOS 6.1
- Leverage common tasks to configure disk and network
- Update README
Change-Id: I7f185de0cfbe3098c9fee2d0f7a792df5b0a95e0