This is not used anymore and py37/38 jobs are broken as it's
not in requierements. Not sure what was installing it earlier.
Closes-Bug: #1995566
Change-Id: I42694d4042574eb1adeb84b66eef637f7f5178a3
(cherry picked from commit aa47e5499cc3d2d84e6317835df9bf49925abd5d)
The tripleo_compute_node is a "meta" role that imports the compute node
standalone playbooks for the purposes of molecule testing. The molecule
default scenario executes the standalone playbooks against a podman in
podman container with systemd as the process, in order to simulate a
compute node.
Change-Id: I14408ad4d2e4fb4ae265a105af18a876158200dc
Signed-off-by: James Slagle <jslagle@redhat.com>
Initial playbooks and inventory directory for deploying a compute node
with only tripleo-ansible. The playbooks will also be consumed from
tripleo-heat-templates when possible, but the primary interface is
entirely ansible-playbook with an inventory and optionally variable
files.
Depends-On: I36fee8e79f71204f3126208d416e8df47a1f6bf8
Change-Id: I96412f44193613d32a8752e8ecfec9324b4fd0fb
Signed-off-by: James Slagle <jslagle@redhat.com>
This change adds a new playbook to configure OVS
for use with ML2/OVN.
This will allow us to leverage Ansible instead of
puppet-vswitch and puppet-ovn on Compute nodes.
Note that this task is currently only targeting
Compute nodes. Controllers would require additional
work to render config files and configure the
DBS bundle + northd.
Change-Id: I36fee8e79f71204f3126208d416e8df47a1f6bf8
Makes the pid namespace configurable for libvirtd containers that use
pid:host. This needs to be configurable for molecule testing when
running podman in podman. Otherwise, pid=host should be used.
Signed-off-by: James Slagle <jslagle@redhat.com>
Change-Id: Ife340f467e5454b588a1c76c8f73858d0c550537
Running the molecule tests locally uses the centos/centos:stream9
images, which does not have systemd installed by default. This patch
adds systemd to the list of packages to install when creating the
container for the molecule default scenario.
This wasn't failing in upstream CI, b/c ubi-9/ubi-init:latest does have
systemd installed by default.
Change-Id: I96643e76f89c4a9b9e556c6ca64517f4f912ba8e
Signed-off-by: James Slagle <jslagle@redhat.com>
currently used puppet-etcd is not a openstack repo, so the
process of updating etcd configuation is complex and time taking.
This patch will introduce a ansible role 'tripleo_etcd' which
replaces puppet-etcd.
This new role generates the etcd configuration based on latest
etcd V3.4
Change-Id: Ia3cf58512c35cd1c2501c606df6fc0f4e4438120
It will make integration into standalone playbooks easier. We also keep
the "main" for backward compatibility reasons.
We also take the opportunity to empty nftables if we're moving from
iptables to nftables; this ensure nothing will interfer.
This action is only done if we switch the firewall engine, meaning,
mostly, during a fresh deploy, or an upgrade. If it happens during an
upgrade, the services using the firewall will already be shut down,
therefore there won't be any weird behavior.
Note: this patch also changes the file extension to match molecule need
to get only .yml, not .yaml. Since the "cleanup.yml" is called in
different places in molecule, this patch must update multiple places at
once.
Change-Id: I761bcf06b01e7890182ec610ca4345c1ecb5707b
There is a script in every node that checks correctly if the
Galera node has been correctly re-introduced in the cluster
as the previous check on the local port open had some
disadvantages, and does not wait for IST to complete.
Change-Id: Id49274f5bade4af5b88b6354cfa092e4b041eb6e
Add design and development docs on how to use the standalone roles and
playbooks.
Document naming rules for standalone role vars, provide examples
and a helper script as the best effort.
Co-authored-by: Bogdan Dobrelya <bdobreli@redhat.com>
Co-authored-by: Chandan Kumar <chkumar@redhat.com>
Depends-On: I14408ad4d2e4fb4ae265a105af18a876158200dc
Change-Id: Ibb585a608c1f8f8192e169b449606a9acb16d338
Signed-off-by: James Slagle <jslagle@redhat.com>
Renames iscsid_install.yaml to install.yml for consistency with
other roles, adds a configure.yml, and renames iscsid.yaml to run.yml.
Since tripleo-heat-templates was already updated to use this role in
e3408da15cbaa070d8d3725dbdece3183cb928b1, symlinks are added for the old
task file names until tripleo-heat-templates is updated.
Also disables managing the service when in check mode, since the needed
registered result would not be available.
Depends-On: https://review.opendev.org/c/openstack/tripleo-heat-templates/+/859999
Change-Id: I64a8541e4388c88d1f3384666eb071c898a9b770
Signed-off-by: James Slagle <jslagle@redhat.com>
While nftables doesn't have any rule de-duplication capabilities, it
doesn't re-create existing chains.
If rules already exist in the "iptables compat layout", reloading the
chains won't drop anything.
Before this patch, if an operator fluches the nftables content (`nft
flush ruleset'), there was a need to manually remove
/etc/nftables/iptables.nft, so that tripleo_nftables could recreate
it, detect its changed state, and load it in nftables.
With this patch, if an operator runs `nft flush ruleset', re-running
the deploy will just re-inject the needed content without any issue.
It also takes care of a misleading comment and task name that were left
untouched after a strategy change. It should therefore be clearer for
future maintenance.
Change-Id: I3d03d7abfd009b724856e6b621867e48e5fafee7
Update tripleo_nova_compute role's jinja2 template to handle
dictionary structure for tripleo_cinder_rbd_multi_config, and
similar, as found in the tripleo_nova_libvirt role.
Change-Id: Ie0800811b63205e327aa8e568c33ac8b829555e7
There are use cases where we'd like to take advantage of the unified
monitoring model in openstack, and the ceph cluster should provide a
set of deployed node exporters, but leave the data collection job to
a different component.
This patch introduces a new condition to allow a partial deployment
of the ceph monitoring stack: only node-exporter(s) are deployed
everywhere, but prometheus/grafana/alertmanager are skipped.
Change-Id: Ia1723394dae5ed03e50f2a236eb808f234e63331
Update default variable handling in tripleo_ceph_client role.
If a variable name used in standalone Ansible roles, like
tripleo_nova_libvirt is used, then tripleo_ceph_client will
support that variable. The following substitutions are now
suported:
tripleo_ceph_client_cluster | tripleo_ceph_cluster_name
tripleo_ceph_client_fsid | tripleo_ceph_cluster_fsid
tripleo_ceph_client_keys | tripleo_ceph_cluster_keys
Also update multiple_external_ceph_clusters.yml so that it
can handle if each 'multiple' item contained in the variable
tripleo_ceph_cluster_multi_config has a "tripleo_ceph_cluster_"
prefix as above. These changes let the role remain backwards
compatible with supported examples in older versions but also
allow the new conventions to be used.
Change-Id: If3f2fde6a40fe3e72b16eee2d93d01aeb773966a
From wallaby we use ephemeral heat and there is no plan to
support upgrade from train to zed. This would remove the
unecessary usage of ansible-collections-openstack.
Change-Id: I67a4623370e66ec486e61ca5eaa2fb33baa8950a
Refactors the role into separate configure.yml and run.yml task
files so that the tasks can be run separately. main.yml includes both
new task files for backwards compatibility.
Change-Id: I48c86a62d897eb3c77935600905d72f256b6b99a
Signed-off-by: James Slagle <jslagle@redhat.com>