Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 1 | .. |
| 2 | SPDX-FileCopyrightText: © 2020 Open Networking Foundation <support@opennetworking.org> |
| 3 | SPDX-License-Identifier: Apache-2.0 |
| 4 | |
| 5 | Site Planning |
| 6 | ============= |
| 7 | |
| 8 | Site Design in Netbox |
| 9 | --------------------- |
| 10 | |
Zack Williams | 2dfb624 | 2021-08-24 09:19:00 -0700 | [diff] [blame] | 11 | The Aether project uses Netbox as source of truth, and the automation script |
| 12 | uses the Netbox API to create input files for Ansible Playbooks which are used |
| 13 | to configure each site. |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 14 | |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 15 | Once the hardware has been ordered, the installation can be planned. The |
| 16 | following information needs to be added to `NetBox |
| 17 | <https://netbox.readthedocs.io/en/stable>`_ to describe each edge site: |
| 18 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 19 | .. note:: |
| 20 | The **bold** words represent the models in Netbox. |
| 21 | |
| 22 | 1. Add a **Site** for the edge (if one doesn't already exist), which has the |
| 23 | physical location and contact information for the Aether Edge. |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 24 | |
| 25 | 2. Add equipment Racks to the Site (if they don't already exist). |
| 26 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 27 | 3. Add a **Tenant** for the edge (who owns/manages it), assigned to the ``Pronto`` |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 28 | or ``Aether`` Tenant Group. |
| 29 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 30 | 4. Add a **VRF** (Routing Table) for the edge site. This is usually just the name |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 31 | of the site. Make sure that ``Enforce unique space`` is checked, so that IP |
| 32 | addresses within the VRF are forced to be unique, and that the Tenant Group |
| 33 | and Tenant are set. |
| 34 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 35 | 5. Add a **VLAN Group** to the edge site, which groups the site's VLANs and |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 36 | requires that they have a unique VLAN number. |
| 37 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 38 | 6. Add **VLANs** for the edge site. These VLAN objects should be assigned a VLAN |
| 39 | Group, a Site, and a Tenant. |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 40 | |
| 41 | There can be multiple of the same VLAN in NetBox (VLANs are layer 2, and |
| 42 | local to the site), but not within the VLAN group. |
| 43 | |
| 44 | The minimal list of VLANs: |
| 45 | |
| 46 | * ADMIN 1 |
| 47 | * UPLINK 10 |
| 48 | * MGMT 800 |
| 49 | * FAB 801 |
| 50 | |
| 51 | If you have multiple deployments at a site using the same management server, |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 52 | add additional VLANs incremented by 10 for the MGMT/FAB - for example, you |
| 53 | can create the VLANs for development server as follows: |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 54 | |
| 55 | * DEVMGMT 810 |
| 56 | * DEVFAB 801 |
| 57 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 58 | 7. Add IP **Prefixes** for the site. This should have the Tenant and VRF assigned. |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 59 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 60 | All IP prefixes of Aether Edge will fit into a ``/22`` sized block. |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 61 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 62 | The Prefix description field is used to create DNS names for IP addresses in the Prefix. |
| 63 | The DNS A records for each IP address start with the name of the Device, and end with |
| 64 | the Prefix description. |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 65 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 66 | For example, if we have a management server named ``mgmtserver`` in **Prefix** |
| 67 | ``prod1.menlo.aetherproject.net``, and the management server's DNS name will be |
| 68 | ``mgmtserver.prod1.menlo.aetherproject.net``. |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 69 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 70 | Here is an example using the ``10.0.0.0/22`` block. Let's name our deployment |
| 71 | as "prod1", and name our site as "menlo". Then we define 4 **Prefixes** |
| 72 | with different purposes. |
| 73 | |
| 74 | .. note:: |
| 75 | NOTE: You should replace the **prod1** and **menlo** to your deployment name and |
| 76 | site name. |
| 77 | .. |
| 78 | |
| 79 | * ADMIN Prefix - ``10.0.0.0/25`` (for Lights-out management) |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 80 | |
| 81 | * Has the Server BMC/LOM and Management Switch |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 82 | * Assign with the ADMIN 1 VLAN |
| 83 | * Set the description to ``admin.prod1.menlo.aetherproject.net`` (or |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 84 | ``prontoproject.net``). |
| 85 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 86 | * MGMT Prefix - ``10.0.0.128/25`` (for infrastructure control plane) |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 87 | |
| 88 | * Has the Server Management plane, Fabric Switch Management/BMC |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 89 | * Assign with MGMT 800 VLAN |
| 90 | * Set the description to ``prod1.menlo.aetherproject.net`` (or |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 91 | ``prontoproject.net``). |
| 92 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 93 | * FABRIC1 Prefix - ``10.0.1.0/25`` |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 94 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 95 | * Compute Nodes' qsfp0 port which connects to Fabric switches, |
| 96 | and other devices (eNB, ...) connect to the Fabric switches. |
| 97 | * Assign with FAB 801 VLAN |
| 98 | * Set the description to ``fab1.prod1.menlo.aetherproject.net`` (or |
| 99 | ``prontoproject.net``). |
| 100 | |
| 101 | * FABRIC2 Prefix - ``10.0.1.128/25`` |
| 102 | |
| 103 | * Compute Nodes' qsfp1 port which connects to Fabric switches |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 104 | * Assign FAB 801 VLAN |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 105 | * Set the description to ``fab2.prod1.menlo.aetherproject.net`` (or |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 106 | ``prontoproject.net``). |
| 107 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 108 | And we will have an additional parent prefix includes 2 FABRIC Prefix. |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 109 | |
| 110 | * ``10.0.1.0/24`` |
| 111 | |
| 112 | * This is used to configure the correct routes, DNS, and TFTP servers |
| 113 | provided by DHCP to the equipment that is connected to the fabric |
| 114 | leaf switch that the management server (which provides those |
| 115 | services) is not connected to. |
| 116 | |
| 117 | Additionally, these edge prefixes are used for Kubernetes but don't need to |
| 118 | be created in NetBox: |
| 119 | |
| 120 | * ``10.0.2.0/24`` |
| 121 | |
| 122 | * Kubernetes Pod IP's |
| 123 | |
| 124 | * ``10.0.3.0/24`` |
| 125 | |
| 126 | * Kubernetes Cluster IP's |
| 127 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 128 | 8. Add **Devices** to the site, for each piece of equipment. These are named with a |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 129 | scheme similar to the DNS names used for the pod, given in this format:: |
| 130 | |
| 131 | <devname>.<deployment>.<site> |
| 132 | |
| 133 | Examples:: |
| 134 | |
| 135 | mgmtserver1.ops1.tucson |
| 136 | node1.stage1.menlo |
| 137 | |
| 138 | Note that these names are transformed into DNS names using the Prefixes, and |
| 139 | may have additional components - ``admin`` or ``fabric`` may be added after |
| 140 | the ``<devname>`` for devices on those networks. |
| 141 | |
| 142 | Set the following fields when creating a device: |
| 143 | |
| 144 | * Site |
| 145 | * Tenant |
| 146 | * Rack & Rack Position |
| 147 | * Serial number |
| 148 | |
| 149 | If a specific Device Type doesn't exist for the device, it must be created, |
| 150 | which is detailed in the NetBox documentation, or ask the OPs team for help. |
| 151 | |
| 152 | See `Rackmount of Equipment`_ below for guidance on how equipment should be |
| 153 | mounted in the Rack. |
| 154 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 155 | 9. Add **Service** to the management server: |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 156 | |
| 157 | * name: ``dns`` |
| 158 | protocol: UDP |
| 159 | port: 53 |
| 160 | |
| 161 | * name: ``tftp`` |
| 162 | protocol: UDP |
| 163 | port: 69 |
| 164 | |
| 165 | These are used by the DHCP and DNS config to know which servers offer |
| 166 | DNS or TFTP service. |
| 167 | |
| 168 | 10. Set the MAC address for the physical interfaces on the device. |
| 169 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 170 | You may also need to add physical network interfaces if they aren't already |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 171 | created by the Device Type. An example would be if additional add-in |
| 172 | network cards were installed. |
| 173 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 174 | 11. Add any virtual interfaces to the **Devices**. When creating a virtual |
| 175 | interface, it should have it's ``label`` field set to the name of the |
| 176 | physical interface that it is assigned |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 177 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 178 | These are needed for two cases of the Pronto deployment: |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 179 | |
| 180 | 1. On the Management Server, there should bet (at least) two VLAN |
| 181 | interfaces created attached to the ``eno2`` network port, which |
| 182 | are used to provide connectivity to the management plane and fabric. |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 183 | These interfaces should be named ``<name of vlan><vlan ID>``, so the |
| 184 | MGMT 800 VLAN would become a virtual interface named ``mgmt800``, with |
| 185 | the label ``eno2``. |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 186 | |
| 187 | 2. On the Fabric switches, the ``eth0`` port is shared between the OpenBMC |
| 188 | interface and the ONIE/ONL installation. Add a ``bmc`` virtual |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 189 | interface with a label of ``eth0`` on each fabric switch, and have the |
| 190 | ``OOB Management`` checkbox checked. |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 191 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 192 | 12. Create **IP addresses** for the physical and virtual interfaces. These should |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 193 | have the Tenant and VRF set. |
| 194 | |
| 195 | The Management Server should always have the first IP address in each |
| 196 | range, and they should be incremental, in this order. Examples are given as |
| 197 | if there was a single instance of each device - adding additional devices |
| 198 | would increment the later IP addresses. |
| 199 | |
| 200 | * Management Server |
| 201 | |
| 202 | * ``eno1`` - site provided public IP address, or blank if DHCP |
| 203 | provided |
| 204 | |
| 205 | * ``eno2`` - 10.0.0.1/25 (first of ADMIN) - set as primary IP |
| 206 | * ``bmc`` - 10.0.0.2/25 (next of ADMIN) |
| 207 | * ``mgmt800`` - 10.0.0.129/25 (first of MGMT) |
| 208 | * ``fab801`` - 10.0.1.1/25 (first of FAB) |
| 209 | |
| 210 | * Management Switch |
| 211 | |
| 212 | * ``gbe1`` - 10.0.0.3/25 (next of ADMIN) - set as primary IP |
| 213 | |
| 214 | * Fabric Switch |
| 215 | |
| 216 | * ``eth0`` - 10.0.0.130/25 (next of MGMT), set as primary IP |
| 217 | * ``bmc`` - 10.0.0.131/25 |
| 218 | |
| 219 | * Compute Server |
| 220 | |
| 221 | * ``eth0`` - 10.0.0.132/25 (next of MGMT), set as primary IP |
| 222 | * ``bmc`` - 10.0.0.4/25 (next of ADMIN) |
| 223 | * ``qsfp0`` - 10.0.1.2/25 (next of FAB) |
| 224 | * ``qsfp1`` - 10.0.1.3/25 |
| 225 | |
| 226 | * Other Fabric devices (eNB, etc.) |
| 227 | |
| 228 | * ``eth0`` or other primary interface - 10.0.1.4/25 (next of FAB) |
| 229 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 230 | 13. Add **IP address** to the **Prefix** to represent reserved DHCP ranges. |
| 231 | We use a single IP address which ``Status`` is set to ``DHCP`` to stand |
| 232 | for the DHCP range, the DHCP server will consume the entire range of IP |
| 233 | address in the CIDR mask (includes first and last IP addresses). |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 234 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 235 | For example, IP ``10.0.0.32/27`` with ``DHCP`` status in Prefix |
| 236 | ``10.0.0.0/25``, the IP will be a DHCP block, and allocate IP address from |
| 237 | ``10.0.0.32`` to ``10.0.0.63``. |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 238 | |
Wei-Yu Chen | f6c0642 | 2021-08-11 11:43:10 +0800 | [diff] [blame] | 239 | 14. Add **IP address** to the **Prefix** to represent route IP reservations for |
| 240 | both Fabric prefixes. These are IP addresses used by ONOS to route traffic |
| 241 | to the other leaf, and have the following attributes: |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 242 | |
| 243 | - Have the last usable address in range (in the ``/25`` fabric examples |
| 244 | above, these would be ``10.0.1.126/25`` and ``10.0.1.254/25``) |
| 245 | |
| 246 | - Have a ``Status`` of ``Reserved``, and the VRF, Tenant Group, and Tenant |
| 247 | set. |
| 248 | |
| 249 | - The Description must start with the word ``router``, such as: ``router |
| 250 | for leaf1 Fabric`` |
| 251 | |
| 252 | - A custom field named ``RFC3442 Routes`` is set to the CIDR IP address of |
| 253 | the opposite leaf - if the leaf's prefix is ``10.0.1.0/25`` and the |
| 254 | router IP is ``10.0.1.126/25`` then ``RFC3442 Routes`` should be set to |
| 255 | ``10.0.1.128\25`` (and the reverse - on ``10.0.1.254/25`` the ``RFC3442 |
| 256 | Routes`` would be set to be ``10.0.1.0/25``). This creates an `RFC3442 |
| 257 | Classless Static Route Option <https://tools.ietf.org/html/rfc3442>`_ |
| 258 | for the subnet in DHCP. |
| 259 | |
| 260 | 15. Add Cables between physical interfaces on the devices |
| 261 | |
| 262 | The topology needs to match the logical diagram presented in the |
Zack Williams | 589ac23 | 2021-08-19 08:52:18 -0700 | [diff] [blame] | 263 | :ref:`Production Environments <edge_deployment/overview:Production |
| 264 | Environments>`. Note that many of the management interfaces need to be |
| 265 | located either on the MGMT or ADMIN VLANs, and the management switch is |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 266 | used to provide that separation. |
| 267 | |
| 268 | Rackmount of Equipment |
| 269 | ---------------------- |
| 270 | |
Zack Williams | 2dfb624 | 2021-08-24 09:19:00 -0700 | [diff] [blame] | 271 | Most of the switch and server equipment used for Aether is available in the 19" |
| 272 | rackmount form factor, but care needs to be taken in mounting this equipment to |
| 273 | ensure proper airflow. Please follow these guidelines: |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 274 | |
| 275 | - The EdgeCore Wedge Switches have a front-to-back (aka "port-to-power") fan |
| 276 | configuration, so hot air exhaust is out the back of the switch near the |
| 277 | power inlets, away from the 32 QSFP network ports on the front of the switch. |
| 278 | |
Zack Williams | 589ac23 | 2021-08-19 08:52:18 -0700 | [diff] [blame] | 279 | - The full-depth servers (such as the 1U and 2U Supermicro servers used in |
| 280 | Pronto) also have front-to-back airflow but have most of their ports on the |
| 281 | rear of the device. |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 282 | |
| 283 | - Airflow through the rack should be in one direction to avoid heat being |
| 284 | pulled from one device into another. This means that to connect the QSFP |
| 285 | network ports from the servers to the switches, cabling should be routed |
| 286 | through the rack from front (switch) to back (server). Empty rack spaces |
| 287 | should be reserved for this purpose. |
| 288 | |
Zack Williams | 589ac23 | 2021-08-19 08:52:18 -0700 | [diff] [blame] | 289 | - The short-depth management switches and management servers should be mounted |
| 290 | on the rear of the rack. They don't generate an appreciable amount of heat, |
| 291 | so the airflow direction isn't a significant factor in racking them. |
Zack Williams | 794532a | 2021-03-18 17:38:36 -0700 | [diff] [blame] | 292 | |
| 293 | Inventory |
| 294 | --------- |
| 295 | |
| 296 | Once equipment arrives, any device needs to be recorded in inventory if it: |
| 297 | |
| 298 | 1. Connects to the network (has a MAC address) |
| 299 | 2. Has a serial number |
| 300 | 3. Isn't a subcomponent (disk, add-in card, linecard, etc.) of a larger device. |
| 301 | |
| 302 | The following information should be recorded for every device: |
| 303 | |
| 304 | - Manufacturer |
| 305 | - Model |
| 306 | - Serial Number |
| 307 | - MAC address (for the primary and any management/BMC/IPMI interfaces) |
| 308 | |
| 309 | This information should be be added to the corresponding Devices within the ONF |
| 310 | NetBox instance. The accuracy of this information is very important as it is |
| 311 | used in bootstrapping the compute systems, which is currently done by Serial |
| 312 | Number, as reported to iPXE by SMBIOS. |
| 313 | |
| 314 | Once inventory has been completed, let the Infra team know, and the pxeboot |
| 315 | configuration will be generated to have the OS preseed files corresponding to the |
| 316 | new servers based on their serial numbers. |