1 NFVBENCH VM IMAGES FOR OPENSTACK
2 ++++++++++++++++++++++++++++++++
4 This repo will build two centos 7 images with:
5 - testpmd and VPP installed for loop VM use case
6 - NFVbench and TRex installed for generator VM use case
8 These VMs will come with a pre-canned user/password: nfvbench/nfvbench
16 - the following packages must be installed prior to using this script:
17 - python3 (+ python3-venv on Ubuntu)
20 - qemu-img (CentOs) or qemu-utils (Ubuntu)
23 .. note:: The image build process is based on `diskimage-builder
24 <https://docs.openstack.org/diskimage-builder/latest/index.html>`_
25 that will be installed in a Python virtual environment by nfvbenchvm
26 build script build-image.sh.
28 .. note:: build-image.sh uses the `gsutil <https://pypi.org/project/gsutil/>`_
29 tool to interact with Google cloud storage (to check if the images
30 exist and to upload the images). This is normally only needed in the
31 context of OPNFV build infrastructure, and build-image.sh can be used
32 without that tool in development environments.
37 - update the version number for the image (if needed) by modifying __version__ in build-image.sh
38 - setup your http_proxy if needed
39 - run ``build-image.sh`` to build the images. A few examples:
41 - to build all the images and publish the code to Google cloud storage:
42 - ``bash build-image.sh``
43 - to build and publish only the loop VM:
44 - ``bash build-image.sh -l``
45 - to build and publish only the generator VM:
46 - ``bash build-image.sh -g``
47 - to build the generator VM without publishing it:
48 - ``bash build-image.sh -gv``
50 .. note:: Run ``bash build-image.sh`` -h to see all options available.
52 .. note:: By default, the generator VM image embeds the latest nfvbench version
53 found on the master branch of OPNFV Gerrit repository
54 https://gerrit.opnfv.org/gerrit/nfvbench.
56 During development phases, it is also possible to build the image with
57 all the committed changes found in the current working copy of
58 nfvbench (local code). To do that, run the image build with the ``-s``
59 option, for instance: ``bash build-image.sh -gvs``.
61 In that case, the version of the generator VM image will be extended
62 with nfvbench development version number to be able to distinguish the
63 development images from the latest published image.
65 LOOP VM IMAGE INSTANCE AND CONFIG
66 =================================
68 Interface Requirements
69 ----------------------
70 The instance must be launched using OpenStack with 2 network interfaces.
71 For best performance, it should use a flavor with:
75 - cpu pinning set to exclusive
79 nfvbench VM will automatically find the two virtual interfaces to use, and use the forwarder specifed in the config file.
81 In the case testpmd is used, testpmd will be launched with mac forwarding mode where the destination macs rewritten according to the config file.
83 In the case VPP is used, VPP will set up a L3 router, and forwarding traffic from one port to the other.
87 nfvbenchvm config file is located at ``/etc/nfvbenchvm.conf``.
92 INTF_MAC1=FA:16:3E:A2:30:41
93 INTF_MAC2=FA:16:3E:10:DA:10
94 TG_MAC1=00:10:94:00:0A:00
95 TG_MAC2=00:11:94:00:0A:00
96 VNF_GATEWAY1_CIDR=1.1.0.2/8
97 VNF_GATEWAY2_CIDR=2.2.0.2/8
100 TG_GATEWAY1_IP=1.1.0.100
101 TG_GATEWAY2_IP=2.2.0.100
104 Launching nfvbenchvm VM
105 -----------------------
107 Normally this image will be used together with NFVBench, and the required configurations will be automatically generated and pushed to VM by NFVBench. If launched manually, no forwarder will be run. Users will have the full control to run either testpmd or VPP via VNC console.
109 To check if testpmd is running, you can run this command in VNC console:
113 sudo screen -r testpmd
115 To check if VPP is running, you can run this command in VNC console:
122 Hardcoded Username and Password
123 --------------------------------
128 GENERATOR IMAGE INSTANCE AND CONFIG
129 ===================================
133 To use openstack APIs, NFVbench generator VM will use `clouds.yaml` file as openstack configuration.
134 The OpenStack clouds configuration from clouds.yaml file to use.
135 clouds.yaml file must be in one of the following paths:
136 - ~/.config/openstack
139 Example of `clouds.yaml`:
146 auth_url: http://192.168.122.10:35357/
150 region_name: RegionOne
152 .. note:: Add `CLOUD_DETAIL` property with the accurate value for your openstack configuration (`devstack` in the above example) in ``/etc/nfvbenchvm.conf``
154 Interface Requirements
155 ----------------------
156 The instance must be launched using OpenStack with 2 network interfaces for dataplane traffic (using SR-IOV function) and 1 management interface to control nfvbench.
157 For best performance, it should use network interfaces for dataplane traffic with a `vnic_type` to `direct-physical` (or `direct` if physical function is not possible)
162 - cpu pinning set to exclusive
164 .. note:: For the management interface: any interface type can be used. This interface required a routable IP (through floating IP or direct) and an access to the openstack APIs.
165 .. note:: CPU pinning: 1 core dedicated for guest OS and NFVbench process, other provided cores are used by TRex
167 Template of a genarator profile using CPU pinning:
175 zmq_pub_port: {{zmq_pub_port}}
176 zmq_rpc_port: {{zmq_rpc_port}}
177 software_mode: {{software_mode}}
180 master_thread_id: '0'
181 latency_thread_id: '1'
184 threads: [{{CORE_THREADS}}]
188 pci: "{{PCI_ADDRESS_1}}"
191 pci: "{{PCI_ADDRESS_2}}"
195 .. note:: `CORE_THREADS` value is determined automatically based on the cores available on the VM starting from 2 to last worker core available.
199 nfvbench VM will automatically find the two virtual interfaces to use for dataplane based on MAC addresses or openstack port name (see config part below).
200 This applies to the management interface as well.
204 nfvbenchvm config file is located at ``/etc/nfvbenchvm.conf``.
206 Example of configuration:
211 LOOPBACK_INTF_MAC1=FA:16:3E:A2:30:41
212 LOOPBACK_INTF_MAC2=FA:16:3E:10:DA:10
213 E2E_INTF_MAC1=FA:16:3E:B0:E2:43
214 E2E_INTF_MAC2=FA:16:3E:D3:6A:FC
216 .. note:: `ACTION` parameter is not mandatory but will permit to start NFVbench with the accurate ports (loopback or e2e).
217 .. note:: Set of MAC parameters cannot be used in parallel as only one NFVbench/TRex process is running.
218 .. note:: Switching from `loopback` to `e2e` action can be done manually using `/nfvbench/start-nfvbench.sh <action>` with the accurate keyword for `action` parameter. This script will restart NFVbench with the good set of MAC.
220 nfvbenchvm config file with management interface:
225 LOOPBACK_INTF_MAC1=FA:16:3E:A2:30:41
226 LOOPBACK_INTF_MAC2=FA:16:3E:10:DA:10
227 INTF_MAC_MGMT=FA:16:3E:06:11:8A
228 INTF_MGMT_CIDR=172.20.56.228/2
229 INTF_MGMT_IP_GW=172.20.56.225
232 .. note:: `INTF_MGMT_IP_GW` and `INTF_MGMT_CIDR` parameters are used by the VM to automatically configure virtual interface and route to allow an external access through SSH.
234 .. note:: ``INTF_MGMT_MTU`` allows to specify the MTU of the management
237 If ``INTF_MGMT_MTU`` is not specified, the MTU will be configured to
238 the conservative value of 1500: this will reduce the risk to get an
241 ``INTF_MGMT_MTU`` can also be set to the special value ``auto``: in
242 that case, the MTU will not be configured and it will keep the value
243 set by the hypervisor (default nfvbench behavior up to version
246 Using pre-created direct-physical ports on openstack, mac addresses value are only known when VM is deployed. In this case, you can pass the port name in config:
250 LOOPBACK_PORT_NAME1=nfvbench-pf1
251 LOOPBACK_PORT_NAME2=nfvbench-pf2
252 E2E_PORT_NAME1=nfvbench-pf1
253 E2E_PORT_NAME1=nfvbench-pf3
254 INTF_MAC_MGMT=FA:16:3E:06:11:8A
255 INTF_MGMT_CIDR=172.20.56.228/2
256 INTF_MGMT_IP_GW=172.20.56.225
257 DNS_SERVERS=8.8.8.8,dns.server.com
259 .. note:: A management interface is required to automatically find the virtual interface to use according to the MAC address provided (see `INTF_MAC_MGMT` parameter).
260 .. note:: NFVbench VM will call openstack API through the management interface to retrieve mac address for these ports
261 .. note:: If openstack API required a host name resolution, add the parameter DNS_SERVERS to add IP or DNS server names (multiple servers can be added separated by a `,`)
263 Control nfvbenchvm VM and run test
264 ----------------------------------
266 By default, NFVbench will be started in server mode (`--server`) and will act as an API.
268 NFVbench VM will be accessible through SSH or HTTP using the management interface IP.
270 NFVbench API endpoint is : `http://<management_ip>:<port>`
271 .. note:: by default port value is 7555
276 To check NFVbench is up and running use REST request:
280 curl -XGET '<management_ip>:<port>/status'
287 "error_message": "nfvbench run still pending",
294 To start a test run using NFVbench API use this type of REST request:
298 curl -XPOST '<management_ip>:<port>/start_run' -H "Content-Type: application/json" -d @nfvbenchconfig.json
300 Example of return when the submission is successful:
305 "error_message": "NFVbench run still pending",
306 "request_id": "42cccb7effdc43caa47f722f0ca8ec96",
311 Start NFVbench test using Xtesting
312 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
314 To start a test run using Xtesting python library and NFVbench API use this type of command on the VM:
318 run_tests -t nfvbench-demo
320 .. note:: `-t` option determine which test case to be runned by Xtesting
321 (see `xtesting/testcases.yaml` file content to see available list of test cases)
324 Connect to the VM using SSH keypair
325 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
327 If a key is provided at VM creation you can use it to log on the VM using `cloud-user` username:
331 ssh -i key.pem cloud-user@<management_ip>
334 Connect to VM using SSH username/password
335 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
337 VM is accessible over SSH using the hardcoded username and password (see below):
341 ssh nfvbench@<management_ip>
344 Launching nfvbenchvm VM
345 -----------------------
347 Normally this image will be deployed using Ansible role, and the required configurations will be automatically generated and pushed to VM by Ansible.
348 If launched manually, users will have the full control to configure and run NFVbench via VNC console.
350 To check if NFVbench is running, you can run this command in VNC console:
354 sudo screen -r nfvbench
357 Hardcoded Username and Password
358 --------------------------------