|
| 1 | +========================= |
| 2 | +Adding and removing hosts |
| 3 | +========================= |
| 4 | + |
| 5 | +This page discusses how to add and remove nodes from an existing cluster. The |
| 6 | +procedure differs depending on the type of nodes being added or removed, which |
| 7 | +services are running, and how they are configured. Here we will consider two |
| 8 | +types of nodes - controllers and compute nodes. Other types of nodes will need |
| 9 | +consideration. |
| 10 | + |
| 11 | +Any procedure being used should be tested before being applied in a production |
| 12 | +environment. |
| 13 | + |
| 14 | +Adding new hosts |
| 15 | +================ |
| 16 | + |
| 17 | +.. _adding-new-controllers: |
| 18 | + |
| 19 | +Adding new controllers |
| 20 | +---------------------- |
| 21 | + |
| 22 | +The :doc:`bootstrap-servers command |
| 23 | +</reference/deployment-and-bootstrapping/bootstrap-servers>` can be used to |
| 24 | +prepare the new hosts that are being added to the system. It adds an entry to |
| 25 | +``/etc/hosts`` for the new hosts, and some services, such as RabbitMQ, require |
| 26 | +entries to exist for all controllers on every controller. If using a |
| 27 | +``--limit`` argument, ensure that all controllers are included, e.g. via |
| 28 | +``--limit control``. Be aware of the :ref:`potential issues <rebootstrapping>` |
| 29 | +with running ``bootstrap-servers`` on an existing system. |
| 30 | + |
| 31 | +.. code-block:: console |
| 32 | +
|
| 33 | + kolla-ansible -i <inventory> bootstrap-servers [ --limit <limit> ] |
| 34 | +
|
| 35 | +Pull down container images to the new hosts. The ``--limit`` argument may be |
| 36 | +used and only needs to include the new hosts. |
| 37 | + |
| 38 | +.. code-block:: console |
| 39 | +
|
| 40 | + kolla-ansible -i <inventory> pull [ --limit <limit> ] |
| 41 | +
|
| 42 | +Deploy containers to the new hosts. If using a ``--limit`` argument, ensure |
| 43 | +that all controllers are included, e.g. via ``--limit control``. |
| 44 | + |
| 45 | +.. code-block:: console |
| 46 | +
|
| 47 | + kolla-ansible -i <inventory> deploy [ --limit <limit> ] |
| 48 | +
|
| 49 | +The new controllers are now deployed. It is recommended to perform testing |
| 50 | +of the control plane at this point to verify that the new controllers are |
| 51 | +functioning correctly. |
| 52 | + |
| 53 | +Some resources may not be automatically balanced onto the new controllers. It |
| 54 | +may be helpful to manually rebalance these resources onto the new controllers. |
| 55 | +Examples include networks hosted by Neutron DHCP agent, and routers hosted by |
| 56 | +Neutron L3 agent. The `removing-existing-controllers`_ section provides an |
| 57 | +example of how to do this. |
| 58 | + |
| 59 | +.. _adding-new-compute-nodes: |
| 60 | + |
| 61 | +Adding new compute nodes |
| 62 | +------------------------ |
| 63 | + |
| 64 | +The :doc:`bootstrap-servers command |
| 65 | +</reference/deployment-and-bootstrapping/bootstrap-servers>`, can be used to |
| 66 | +prepare the new hosts that are being added to the system. Be aware of the |
| 67 | +:ref:`potential issues <rebootstrapping>` with running ``bootstrap-servers`` on |
| 68 | +an existing system. |
| 69 | + |
| 70 | +.. code-block:: console |
| 71 | +
|
| 72 | + kolla-ansible -i <inventory> bootstrap-servers [ --limit <limit> ] |
| 73 | +
|
| 74 | +Pull down container images to the new hosts. The ``--limit`` argument may be |
| 75 | +used and only needs to include the new hosts. |
| 76 | + |
| 77 | +.. code-block:: console |
| 78 | +
|
| 79 | + kolla-ansible -i <inventory> pull [ --limit <limit> ] |
| 80 | +
|
| 81 | +Deploy containers on the new hosts. The ``--limit`` argument may be used and |
| 82 | +only needs to include the new hosts. |
| 83 | + |
| 84 | +.. code-block:: console |
| 85 | +
|
| 86 | + kolla-ansible -i <inventory> deploy [ --limit <limit> ] |
| 87 | +
|
| 88 | +The new compute nodes are now deployed. It is recommended to perform |
| 89 | +testing of the compute nodes at this point to verify that they are functioning |
| 90 | +correctly. |
| 91 | + |
| 92 | +Server instances are not automatically balanced onto the new compute nodes. It |
| 93 | +may be helpful to live migrate some server instances onto the new hosts. |
| 94 | + |
| 95 | +.. code-block:: console |
| 96 | +
|
| 97 | + openstack server migrate <server> --live-migration --host <target host> --os-compute-api-version 2.30 |
| 98 | +
|
| 99 | +Alternatively, a service such as :watcher-doc:`Watcher </>` may be used to do |
| 100 | +this automatically. |
| 101 | + |
| 102 | +Removing existing hosts |
| 103 | +======================= |
| 104 | + |
| 105 | +.. _removing-existing-controllers: |
| 106 | + |
| 107 | +Removing existing controllers |
| 108 | +----------------------------- |
| 109 | + |
| 110 | +When removing controllers or other hosts running clustered services, consider |
| 111 | +whether enough hosts remain in the cluster to form a quorum. For example, in a |
| 112 | +system with 3 controllers, only one should be removed at a time. Consider also |
| 113 | +the effect this will have on redundancy. |
| 114 | + |
| 115 | +Before removing existing controllers from a cluster, it is recommended to move |
| 116 | +resources they are hosting. Here we will cover networks hosted by Neutron DHCP |
| 117 | +agent and routers hosted by Neutron L3 agent. Other actions may be necessary, |
| 118 | +depending on your environment and configuration. |
| 119 | + |
| 120 | +For each host being removed, find Neutron routers on that host and move them. |
| 121 | +Disable the L3 agent. For example: |
| 122 | + |
| 123 | +.. code-block:: console |
| 124 | +
|
| 125 | + l3_id=$(openstack network agent list --host <host> --agent-type l3 -f value -c ID) |
| 126 | + target_l3_id=$(openstack network agent list --host <target host> --agent-type l3 -f value -c ID) |
| 127 | + openstack router list --agent $l3_id -f value -c ID | while read router; do |
| 128 | + openstack network agent remove router $l3_id $router --l3 |
| 129 | + openstack network agent add router $target_l3_id $router --l3 |
| 130 | + done |
| 131 | + openstack network agent set $l3_id --disable |
| 132 | +
|
| 133 | +Repeat for DHCP agents: |
| 134 | + |
| 135 | +.. code-block:: console |
| 136 | +
|
| 137 | + dhcp_id=$(openstack network agent list --host <host> --agent-type dhcp -f value -c ID) |
| 138 | + target_dhcp_id=$(openstack network agent list --host <target host> --agent-type dhcp -f value -c ID) |
| 139 | + openstack network list --agent $dhcp_id -f value -c ID | while read network; do |
| 140 | + openstack network agent remove network $dhcp_id $network --dhcp |
| 141 | + openstack network agent add network $target_dhcp_id $network --dhcp |
| 142 | + done |
| 143 | +
|
| 144 | +Stop all services running on the hosts being removed: |
| 145 | + |
| 146 | +.. code-block:: console |
| 147 | +
|
| 148 | + kolla-ansible -i <inventory> stop --yes-i-really-really-mean-it [ --limit <limit> ] |
| 149 | +
|
| 150 | +Remove the hosts from the Ansible inventory. |
| 151 | + |
| 152 | +Reconfigure the remaining controllers to update the membership of clusters such |
| 153 | +as MariaDB and RabbitMQ. Use a suitable limit, such as ``--limit control``. |
| 154 | + |
| 155 | +.. code-block:: console |
| 156 | +
|
| 157 | + kolla-ansible -i <inventory> deploy [ --limit <limit> ] |
| 158 | +
|
| 159 | +Perform testing to verify that the remaining cluster hosts are operating |
| 160 | +correctly. |
| 161 | + |
| 162 | +For each host, clean up its services: |
| 163 | + |
| 164 | +.. code-block:: console |
| 165 | +
|
| 166 | + openstack network agent list --host <host> -f value -c ID | while read id; do |
| 167 | + openstack network agent delete $id |
| 168 | + done |
| 169 | +
|
| 170 | + openstack compute service list --os-compute-api-version 2.53 --host <host> -f value -c ID | while read id; do |
| 171 | + openstack compute service delete --os-compute-api-version 2.53 $id |
| 172 | + done |
| 173 | +
|
| 174 | +.. _removing-existing-compute-nodes: |
| 175 | + |
| 176 | +Removing existing compute nodes |
| 177 | +------------------------------- |
| 178 | + |
| 179 | +When removing compute nodes from a system, consider whether there is capacity |
| 180 | +to host the running workload on the remaining compute nodes. Include overhead |
| 181 | +for failures that may occur. |
| 182 | + |
| 183 | +Before removing compute nodes from a system, it is recommended to migrate or |
| 184 | +destroy any instances that they are hosting. |
| 185 | + |
| 186 | +For each host, disable the compute service to ensure that no new instances are |
| 187 | +scheduled to it. |
| 188 | + |
| 189 | +.. code-block:: console |
| 190 | +
|
| 191 | + openstack compute service set <host> nova-compute --disable |
| 192 | +
|
| 193 | +If possible, live migrate instances to another host. |
| 194 | + |
| 195 | +.. code-block:: console |
| 196 | +
|
| 197 | + openstack server list --host <host> -f value -c ID | while read server; do |
| 198 | + openstack server migrate --live-migration $server |
| 199 | + done |
| 200 | +
|
| 201 | +Verify that the migrations were successful. |
| 202 | + |
| 203 | +Stop all services running on the hosts being removed: |
| 204 | + |
| 205 | +.. code-block:: console |
| 206 | +
|
| 207 | + kolla-ansible -i <inventory> stop --yes-i-really-really-mean-it [ --limit <limit> ] |
| 208 | +
|
| 209 | +Remove the hosts from the Ansible inventory. |
| 210 | + |
| 211 | +Perform testing to verify that the remaining cluster hosts are operating |
| 212 | +correctly. |
| 213 | + |
| 214 | +For each host, clean up its services: |
| 215 | + |
| 216 | +.. code-block:: console |
| 217 | +
|
| 218 | + openstack network agent list --host <host> -f value -c ID | while read id; do |
| 219 | + openstack network agent delete $id |
| 220 | + done |
| 221 | +
|
| 222 | + openstack compute service list --os-compute-api-version 2.53 --host <host> -f value -c ID | while read id; do |
| 223 | + openstack compute service delete --os-compute-api-version 2.53 $id |
| 224 | + done |
0 commit comments