Id093613f9d410eb3fe5564a724c0f75275eeb4e8 defines a new internal
glance-api service, but the patch contains an error in defining the
network associated with the service. This patch corrects the error.
Closes-Bug: #1998227
Change-Id: I456b4235242cae125f5ad4cd9cc7415f2699462c
This patch adopts the recommendation outlined in OSSN-0090 [1], in
which two instances of the glance-api service are deployed:
- A "user facing" glance-api service, accessible via the Public
keystone endpoint.
- An "internal facing only" service, accessible via the Admin and
Internal keystone endpoints.
The user facing instance is configured so it does not report any image
location information. This is achieved by configuring glance-api.conf
with the show_image_direct_url and show_multiple_locations set to False.
The internal service operates on a separate TCP port (defaults to 9293)
with its own glance-api.conf that configures show_image_direct_url and
show_multiple_locations set to True.
In order for cinder and nova to have access to the image location data,
both services are configured to access glance via the internal service.
[1] https://wiki.openstack.org/wiki/OSSN/OSSN-0090
Closes-Bug: #1822540
Depends-On: Ideb5a951d538d9e2c7cca11dfe0e8b99520de959
Depends-On: Ib6188505197d0a267dbd8c4d96f12f31f7b5c9f0
Change-Id: Id093613f9d410eb3fe5564a724c0f75275eeb4e8
VMAX was rebranded to PowerMax and the old driver was deprecated.
Because current TripleO supports the new PowerMax share driver and
the configurations for old VMAX driver is known to be broken, this
removes support for the VMAX share driver.
Change-Id: I8031f6c66dd1feae303e7b2c3eeddfb8b532c676
These services were deprecated during Xena cycle by [1]. Since we have
created the Zed release, now we can remove the features deprecated
before that.
[1] 18651160a9
Change-Id: Ib248ec4bec57a5d747a1cf35157084d2cb51b566
This patch adds support for IBM Spectrum virtualize family (Svf) driver.
With this new tht for Svf, there is no need to use tripleo's ExtraConfig
technique to configure the backend.
Depends-On: I3c74f492d9d34eb8e16d9a203b5f02714b02f69a
Change-Id: I59c38368356595254e8617ef30157040fc204f92
Currently, Cinder RBD storage backend is configured as part of
cinder-volume-container-puppet.yaml.
This change will just relocate the existing cinder RBD parameters in
their own template. There are no new parameters, and the existing parameters
retain the same default values
Change-Id: I73008a3ff38bafa987f5c2be6fb34d2d110bbaef
Neutron resources such as port should not be created as part of stack
since we replaced undercloud heat service by ephemeral heat. This
change replaces the default type mapping of ControlPlaneVipPort to
avoid using neutron resources by default.
Change-Id: Ida6dc27c67b9a7c3c1af60c1adbff4bb51e40659
The following resource types are no longer used since we removed
software deployment by heat in favor of the config-download process.
- OS::TripleO::WorkflowSteps
- OS::TripleO::DeploymentSteps
Change-Id: I61b9a291e71158eee9f2411711e69d11bea4249e
FFUv3 (Train->Wallaby) and mixed-rhel stories require possibility to
run modular libvirt on el9 hypervisors and monolithic libvirt on el8.
It is not clear yet if we might need a new ComputeEL8 role, or not.
Create legacy service OS::TripleO::Services::NovaLibvirtLegacy to run
monolithic libvirt on such computes that are supposed to stay on
RHEL8.4 (el8) for some time.
Related: rhbz#2136163
Signed-off-by: Bogdan Dobrelya <bdobreli@redhat.com>
Change-Id: I0476851066a5c3e22e09ff02ed4bb19a7eff71a7
This is the prep-work to drop the deployed-server-environment.yaml and
makes sure the deployed server is used by default.
Note that this change also makes the ControlPlanePort resource type
default to the deployed neutron port, because the port should be
pre-provisioned as part of baremetal node deployment process.
Change-Id: I03500eb2b4a6302f35aa71ae2f81bfd0c0bf064f
Now overcloud is always deployed using pre-provisioned nodes, and all
deployment types(standalone, undercloud and overcloud) enables package
installation.
Thus we can update the default value of EnablePackageInstall instead of
setting the parameter everytime.
Change-Id: I5598f6d8ee82f0df2598dd0e48fbc16cdb426009
... since OpenDaylight support was removed by [1].
This also cleans up a few remaining references in environment files.
[1] 9c3638b8a1
Change-Id: Ic01100b74acdf9ba9490aaa8a307580f40f3e032
This change introduces a new implementation to manage CA certificate
files by ansible, which replaces the existing one by puppet.
Depends-on: https://review.opendev.org/743620
Change-Id: I241f3d635e36a1497a1146fdd4c1db7dfde28dc8
The snmp service was first introduced so that undercloud ceilometer can
gather metrics via the snmpd daemon in overcloud nodes.
However Telemetry services are already removed from undercloud, and
this requirement is no longer valid.
This deprecates the snmp service support so that we can remove it in
a future release. This also introduces upgrade tasks to clean up
the service because it has been enabled by default.
Note that the net-snmp package will be removed from the overcloud image,
and using the disabled snmpd service requires adding snmpd package into
the overcloud image or setting up a package repository.
Change-Id: I9d13fccb5d4d265cfc04df5ccb287b6215e7d0b2
Currently, Cinder NFS storage backend is configured as part of
cinder-volume-container-puppet.yaml.
This change will just relocate the existing cinder NFS parameters in their
own template. There are no new parameters, and the existing parameters
retain the same default values.
Change-Id: I10364adb761cc225173cc437cf2876ed7e896f19
Currently ovn is integrated to neutron as a ml2 plugin, not a core
plugin, and this service resource has never been used.
Change-Id: I2433f5001b9fbe547907080e7da8f35c8c2596f8
... because NSX-T 3.x release series is the final series to support KVM
and non-VIO OpenStack distributions and integration is no longer
supported with the recent OpenStack releases.
This directly removes the feature because it can't be used any longer.
Change-Id: I778d9438e3ab1eaf2e940ae9be923c86797f75ee
Default to deploying OVN with RAFT Active/Active clustering instead
of Pacemaker Active/Backup. The code to support RAFT has been been
in THT since train.
Requires OVS 2.17 (technically only python-ovs 2.17) to work
properly at scale.
Change-Id: I535087bc5f1266d1f4fe8bb1dca7d8c355c9de9d
Having rndc on in the internal API network talking to bind instances on
the external network creates an asymmetric routing situation that
requires loosening the reverse path filter. This moves all rndc traffic
onto the same network instead, removing the asymmetry and the need for
loosening the filter.`
Change-Id: Ieb7f39406ace080f1c136441add26a5a167f0d57
Support for xinetd service was deprecated long ago[1] and actual
implementation was removed at that time.
This change removes the remaining service type.
[1] 78f1901da4
Change-Id: I2906e37991443ecacb1557de1618bb5f8f135b1f
This change removes the following parameters, which were used by
undercloud Nova.
- KeyName
- Overcloud{{role.name}}Flavor
- {{role.name}}SchedulerHints
- {{role.name}}Image
This also removes the NodeUserData resource because it depends on
cloud-init and nova metadata and is no longer used since Nova was
removed from baremetal node provisioning.
Finally, this change makes deployed server method used by default, and
removes remaining implementation to keep the resource compatible with
OS::Nova::Server.
Change-Id: I571b401ab2ca3c77352f4849eb2b99de20292032