Age | Commit message (Collapse) | Author | Files | Lines |
|
|
|
|
|
|
|
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
|
|
|
|
|
|
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
|
|
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
|
|
|
|
|
|
Keystone doesn't provide different flags to indicate that both of its
endpoints are enabled. So currently we have to manually add its
network to all-nodes-config.
bp tls-via-certmogner
Change-Id: Ibecd78706e84853107f698ba411a0c05e6f5be52
|
|
|
|
Include the neutron-base service definition to align pull in common
configurations. This might not be *absolutely* necessary as any required
common configuration would likely already be added by and OVS agent
service, etc. but it's better to be safe than sorry and it does keep
things consistent across the Neutron services.
Implements: blueprint tripleo-sriov
Change-Id: I10a9d9b29760475e6cd75e4057051c75a52ffbb7
|
|
This will aid us in using FQDNs instead of IPs if DNS is not set. If
the deployer already has DNS set up, they can easily disable this
profile by adding the use-dns-for-vips.yaml environment file.
bp tls-via-certmonger
Change-Id: I8c1b3f253d0149d575171c208f9a1342a7b26450
Depends-On: I1bdb2701dfb3e7ef072e674c9882d3be5af7296c
|
|
The nodes need to be aware of the fqdn's for the specific endpoints
in the cloud. This could be either to set the entries in /etc/hosts
or to select an appropriate hostname for a certificate to be
generated.
bp tls-via-certmonger
Change-Id: I9b4645b937a344f46ec18a9a68c5afa2bc5206d0
|
|
We recently made changes that add data to allNodesConfig, but
we didn't wire the files into the hierarchy on all roles.
Change-Id: I8e838b02bd982e600af54b14350106322244890a
Closes-Bug: #1620485
|
|
|
|
|
|
|
|
|
|
|
|
puppet aodh auth type defaults to password type and v2
auth_url doesnt work with domain. This fixes the url to
not include suffix.
Change-Id: I46d53e748d8932ed1183bedbdeb5eefcde679f9e
|
|
Previously we weren't creating Redis VIP in keepalived, causing Redis to
be unusable in non-HA deployments. This is now fixed.
Depends-On: I0bb37f6fb3eed022288b2dcfc7a88e8ff88a7ace
Change-Id: I0ecfda1e6ad5567f6f58d60bf418bc91761833ab
Closes-Bug: #1618510
|
|
Move Redis VIP from controller-only to all nodes so that we don't assume
where Redis is deployed.
Change-Id: I55f8d48e3e077951fbcc88158dd6f21a2fe5f457
Related-Bug: #1618510
Partially-Implements: blueprint custom-roles
|
|
This adds a mapping of which service is on which network. This
information can be used to fetch a certificate depending on the
network (since they use different hostnames).
Change-Id: I176245da591bea28aeabf3d2b552f24456c98c43
|
|
|
|
|
|
|
|
|
|
|
|
This makes it easier to access the VIP data for other node types and
de-ties this from the controller role.
Change-Id: I71125576ec93889fed134b92fb59f7e7dc9920c4
|
|
The goal is to have only production drivers available out of box.
Remove agent_ipmitool as it required additional configuration of
swift temporary URLs. Add pxe_ilo and pxe_drac as they work out of box.
This brings the default enabled drivers list closer to one of undercloud.
Note that pxe_ssh is not enabled as it's not a production driver.
Change-Id: Ic7d2190e2d39a795b024c2999e22ee47b3045cf1
Partially-Implements: blueprint ironic-integration
|
|
Moved this list manipulation to puppet-tripleo where it can easily
reference the memcached_node_ips array instead of hard-coded output
from the Controller.
Depends-On: Ic8872e5e51732874ca5b93bff5efd3e7ed75bc31
Change-Id: I523a802c8d22be7602728ebe8966d9e2ee109cde
Partially-Implements: blueprint custom-roles
|
|
Change-Id: Iaf78a5eac223aeb50133fcc6fab043ee746d73fd
Partially-Implements: blueprint custom-roles
|
|
To avoid the hard-coded references which won't work with
composable roles, we instead default to the rabbitmq_node_ips
list in the per-service puppet-tripleo profiles.
Change-Id: I76b7e06781fdd5d969503b6d73423bb3f5f7a41f
Depends-On: Ie53c93456529420588eb1927703ea91b54095d87
Partially-Implements: blueprint custom-roles
|
|
|
|
|
|
This patch moves settings for pacemaker and the tripleo firewall
out of controller.yaml.
Related bug: #1604414
Change-Id: I0164717bfd79cdea3de8eb7a64771028bea201ac
|
|
|
|
|
|
Some puppet interfaces require a comma separated list of hostnames
where a service is running, so generate it in a similar way to th
service ips.
Change-Id: Icdf5d993d089dc94035194bdbd52299fcbc793be
Partially-Implements: blueprint custom-roles
|
|
|
|
Provides snmp_readonly_user_* parameters to the base::snmp class
and removes them from the role templates.
Related-bug: #1604414
Co-Authored-By: Giulio Fidente <gfidente@redhat.com>
Change-Id: I35b1ff23baff0c25e742562e1103a3882f7bf767
|
|
Cleaning is run before the first and between every deployments to ensure
that the node is in a clean state (hence its name). The most important
part of it is wiping the hard drive.
This change allows tuning this action. A new option IronicCleaningDiskErase
accepts the following options:
* full - for full cleaning with removing the data via ATA secure erase or shred
* metadata - for cleaning only partitioning table.
The third option "none" is suppored by Puppet but is not advertised here as
disabling disk wiping completely is known to cause bugs.
Note that the upstream Ironic default is a bit unfortunate: it enabled both
processes. We set the default to "full" which has the same effect, but is faster.
Change-Id: I46632f71f7ab63daac243acefa5166c58ca05da0
Partially-Implements: blueprint ironic-integration
|
|
Introduce two new options: IronicIPXEEnabled (true by default) and
IronicIPXEPort (8088 by default).
Also copy UEFI configuration from instack-undercloud.
Change-Id: I4e037c1ea9c8663efb4d472469d0d2640e265710
Depends-On: Icb9633134114041bbd497e7652482dd5d34c9327
Partially-Implements: blueprint ironic-integration
|
|
Change-Id: I30003e4500ccf681121bca7f0a8ffddb05f61660
Partially-Implements: blueprint ironic-integration
|
|
|
|
Installs and configures OVN databases and process and also configures
the neutron plugin for OVN.
Implements: blueprint tripleo-ovn
Change-Id: Ic3e415bb0587dd85b71f6c14a96f6b2c86a7b30f
|
|
Neutron's behavior changed in Newton with respect to the default value
of 0 for NeutronWorkers. Instead of effectively treating it as
"processorcount", it now assumes we mean what we say and doesn't spawn
any workers, resulting in only a single process for handling API
requests. This change alters the default value to regain the previous
behavior.
This change also extends the worker count to the RPC worker setting -
which was possibly an oversight in previous releases. The default
behavior will have no result for most systems.
Closes-Bug: 1619383
Change-Id: Id6e3ee544160372c806905e77348150a1a1eec1b
|
|
|