@ -53,7 +53,7 @@ Preparing the Baremetal Environment
Networking
^^^^^^^^^^
The overcloud nodes will be deployed from the undercloud machine and therefore the machines need to have their network settings modified to allow for the overcloud nodes to be PXE boot'ed using the undercloud machine. As such, the setup requires that:
The overcloud nodes will be deployed from the undercloud machine and therefore the machines need to have their network settings modified to allow for the overcloud nodes to be PXE booted using the undercloud machine. As such, the setup requires that:
* All overcloud machines in the setup must support IPMI
* A management provisioning network is setup for all of the overcloud machines.
The following example will peform following steps:
The following example will perform following steps:
* ensure **ndctl** tool is installed on hosts with role **ComputePMEM**
* create PMEM namespaces as specified in the **NovaPMEMNamespaces** parameter.
- ns0, ns1, ns2 with size 6GiB
- ns3 with size 100GiB
* set Nova prameter **pmem_namespaces** in nova.conf to map create namespaces to vPMEM as specified in **NovaPMEMMappings**.
* set Nova parameter **pmem_namespaces** in nova.conf to map create namespaces to vPMEM as specified in **NovaPMEMMappings**.
In this example the label '6GB' will map to one of ns0, ns1 or ns2 namespace and the label 'LARGE' will map to ns3 namespace.
After deployment you need to configure flavors as described in documentation `Nova: Configure a flavor <https://docs.openstack.org/nova/latest/admin/virtual-persistent-memory.html#configure-a-flavor>`_
There is an ansible project called opstools-ansible (OpsTools_) on github that helps to install the Operator Server, further documentation of the operational tool server instalation can be founded at (OpsToolsDoc_).
There is an ansible project called opstools-ansible (OpsTools_) on github that helps to install the Operator Server, further documentation of the operational tool server installation can be founded at (OpsToolsDoc_).
@ -17,7 +17,7 @@ where instance creation times must be minimized.
Since Ussuri Nova also provides an API to pre-cache images on Compute nodes.
See the `Nova Image pre-caching documentation <https://docs.openstack.org/nova/ussuri/admin/image-caching.html#image-pre-caching>`_.
..note:: The Nova Image Cache is not used when using Ceph RBD for Glange images and Nova ephemeral disk. See `Nova Image Caching documentation <https://docs.openstack.org/nova/ussuri/admin/image-caching.html>`_.
..note:: The Nova Image Cache is not used when using Ceph RBD for Glance images and Nova ephemeral disk. See `Nova Image Caching documentation <https://docs.openstack.org/nova/ussuri/admin/image-caching.html>`_.
Image Cache Cleanup
-------------------
@ -167,7 +167,7 @@ Pre-caching on one node and distributing to remaining nodes
In the case of a :doc:`../features/distributed_compute_node` it may be desirable to transfer an image to a single compute node at a remote site and then redistribute it from that node to the remaining compute nodes.
The SSH/SCP configuration that exists between the compute nodes to support cold migration/resize is reused for this purpose.
..warning:: SSH/SCP is inefficient over high latency networks. The method should only be used when the compute nodes targetted by the playbook are all within the same site. To ensure this is the case set tripleo_nova_image_cache_plan to the stack name of the site. Multiple runs of ansible-playbook are then required, targetting a different site each time.
..warning:: SSH/SCP is inefficient over high latency networks. The method should only be used when the compute nodes targeted by the playbook are all within the same site. To ensure this is the case set tripleo_nova_image_cache_plan to the stack name of the site. Multiple runs of ansible-playbook are then required, targeting a different site each time.
To enable this simply set `tripleo_nova_image_cache_use_proxy: true` in the arguments file.
The image is distributed from the first compute node by default. To use a specific compute node also set `tripleo_nova_image_cache_proxy_hostname`.
Zuul
3 months ago
committed by