starlingx/docs
Code Issues Proposed changes

Update Docs to include Hybrid Cluster (r6)

Browse Source 
Add documentation about "hybrid clusters" where there are a mix of Kubernetes and OpenStack nodes in the same cluster.
Add documentation about converting worker nodes to compute nodes.
Removed new content from r5 as it was not relevant.
Fixed typos.
New step added in Convert Worker Nodes section.

Signed-off-by: Elisamara Aoki Goncalves <elisamaraaoki.goncalves@windriver.com>
Change-Id: Ie39bc8b94e2281649a26b13ceb778325ff53422e
This commit is contained in:
Elisamara Aoki Goncalves 2022-04-13 22:35:37 -03:00
parent 8687488811
commit 579a08d6fd
4 changed files with 195 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

36
doc/source/deploy_install_guides/r6_release/openstack/add-taints-to-openstack-node-in-hybrid-cluster-e8b37e8d1b48.rst Normal file
View File

@ -0,0 +1,36 @@
.. _add-taints-to-openstack-node-in-hybrid-cluster-e8b37e8d1b48-r6:
==============================================
Add Taints to OpenStack Node in Hybrid Cluster
==============================================
.. rubric:: |context|
In a hybrid (Kubernetes and OpenStack) cluster scenario, to prevent end-users'
hosted containerized workloads/pods from being scheduled on
``openstack-compute-nodes`` a taint that only openstack and platform pods are
able to tolerate must be added. Thus, any pod that is not platform or
OpenStack specific can be repelled. To achieve this control, the
``openstack-compute-node`` taint must be added to all
``openstack-compute-nodes`` (i.e. worker nodes or |AIO|-Controller nodes with
the ``openstack-compute-node`` label).
By applying taints, it is possible to separate end users' containerized
workloads/pods from OpenStack in a hybrid (Kubernetes and OpenStack) cluster
scenario.
.. rubric:: |proc|
#. Apply taints.
This step is needed to prevent end-users' hosted containerized
workloads/pods from being scheduled on openstack computes (i.e. worker
nodes or |AIO|-Controller nodes with the ``openstack-compute-node`` label).
Apply this taint on every ``openstack-compute-node`` (i.e. worker nodes
or |AIO|-Controller nodes with the ``openstack-compute-node`` label):
.. code-block:: none
kubectl taint nodes <kubernetes-node-name> openstack-compute-node:NoSchedule

103
doc/source/deploy_install_guides/r6_release/openstack/convert-worker-nodes-0007b1532308.rst Normal file
View File

@ -0,0 +1,103 @@
.. _convert-worker-nodes-0007b1532308:
====================
Convert Worker Nodes
====================
.. rubric:: |context|
In a hybrid (Kubernetes and OpenStack) cluster scenario you may need to convert
worker nodes to/from ``openstack-compute-nodes``.
.. rubric:: |proc|
#. Convert a k8s-only worker into a OpenStack compute
#. Lock the worker host:
.. code-block:: none
system host-lock <host>
#. Add the ``openstack-compute-node`` taint, see :ref:`Add Taints to
OpenStack Node in Hybrid Cluster
<add-taints-to-openstack-node-in-hybrid-cluster-e8b37e8d1b48-r6>`.
#. Assign OpenStack labels:
.. code-block:: none
system host-label-assign <host> --overwrite openstack-compute-node=enabled avs=enabled sriov=enabled
#. Allocate vswitch huge pages:
.. code-block:: none
system host-memory-modify -1G 1 -f vswitch <host> 0
system host-memory-modify -1G 1 -f vswitch <host> 1
#. Change the class of the data network interface:
.. code-block:: none
system host-if-modify -c data <host> <if_name_or_uuid>
.. note::
If data network interface does not exist yet, refer to |prod-os|
documentation on creating it.
#. Change Kubernetes CPU Manager Policy to allow |VMs| to use application
cores:
.. code-block:: none
system host-label-remove <host> kube-cpu-mgr-policy
#. Unlock the worker host:
.. code-block:: none
system host-unlock <host>
#. Convert a OpenStack compute into a k8s-only worker.
#. Lock the worker host:
.. code-block:: none
system host-lock <host>
#. Remove OpenStack labels:
.. code-block:: none
system host-label-remove <host> openstack-compute-node avs sriov
.. note::
The labels have to be removed, not to have its values changed.
#. Deallocate vswitch huge pages:
.. code-block:: none
system host-memory-modify -1G 0 -f vswitch <host> 0
system host-memory-modify -1G 0 -f vswitch <host> 1
#. Change the class of the data network interface:
.. code-block:: none
system host-if-modify -c none <host> <if_name_or_uuid>
.. note::
This change is needed to avoid raising a permanent alarm for the
interface without the need to delete it.
#. Unlock the worker host:
.. code-block:: none
system host-unlock <host>

45
doc/source/deploy_install_guides/r6_release/openstack/hybrid-cluster-c7a3134b6f2a.rst Normal file
View File

@ -0,0 +1,45 @@
.. _hybrid-cluster-c7a3134b6f2a:
==============
Hybrid Cluster
==============
A Hybrid Cluster occurs when the hosts with a worker function (|AIO|
controllers and worker nodes) are split between two groups, one running
|prod-os| for hosting |VM| payloads and the other for hosting containerized
payloads.
The host labels are used to define each worker function on the Hybrid Cluster
setup. For example, a standard configuration (2 controllers and 2 computes) can
be split into (2 controllers, 1 openstack-compute and 1 kubernetes-worker).
-----------
Limitations
-----------
- Worker function on |AIO| controllers MUST both be either
Kubernetes or OpenStack.
- Hybrid Cluster does not apply to |AIO-SX| or |AIO-DX| setups.
- A worker must have only one function, either it is OpenStack compute or
k8s-only worker, never both at the same time.
- The ``sriov`` and ``sriovdp`` labels cannot coexist on the same host,
in order to prevent the |SRIOV| device plugin from conflicting with the
OpenStack |SRIOV| driver.
- No host will assign |VMs| and application containers to application cores
at the same time.
- Standard Controllers cannot have ``openstack-compute-node`` label;
only |AIO| Controllers can have ``openstack-compute-node`` label.
- Taints must be added to OpenStack compute hosts (i.e. worker nodes or
|AIO|-Controller nodes with the ``openstack-compute-node`` label) to
prevent
end users' hosted containerized workloads/pods from being scheduled on
OpenStack compute hosts. See :ref:`Add Taints to OpenStack Node in
Hybrid Cluster
<add-taints-to-openstack-node-in-hybrid-cluster-e8b37e8d1b48-r6>`.

11
doc/source/deploy_install_guides/r6_release/openstack/index-install-r6-os-adc44604968c.rst
View File

@ -16,3 +16,14 @@ StarlingX is independent of deployment configuration.
install
access
uninstall_delete
--------------
Hybrid Cluster
--------------
.. toctree::
:maxdepth: 1
hybrid-cluster-c7a3134b6f2a
add-taints-to-openstack-node-in-hybrid-cluster-e8b37e8d1b48
convert-worker-nodes-0007b1532308