docs/testing/user/userguide/opnfv_yardstick_tc053.rst

   1 .. This work is licensed under a Creative Commons Attribution 4.0 International
   2 .. License.
   3 .. http://creativecommons.org/licenses/by/4.0
   4 .. (c) OPNFV, Yin Kanglin and others.
   5 .. 14_ykl@tongji.edu.cn
   6
   7 *************************************
   8 Yardstick Test Case Description TC053
   9 *************************************
  10
  11 +-----------------------------------------------------------------------------+
  12 |OpenStack Controller Load Balance Service High Availability                  |
  13 |                                                                             |
  14 +--------------+--------------------------------------------------------------+
  15 |test case id  | OPNFV_YARDSTICK_TC053: OpenStack Controller Load Balance     |
  16 |              | Service High Availability                                    |
  17 +--------------+--------------------------------------------------------------+
  18 |test purpose  | This test case will verify the high availability of the      |
  19 |              | load balance service(current is HAProxy) that supports       |
  20 |              | OpenStack on controller node. When the load balance service  |
  21 |              | of a specified controller node is killed, whether other load |
  22 |              | balancers on other controller nodes will work, and whether   |
  23 |              | the controller node will restart the load balancer are       |
  24 |              | checked.                                                     |
  25 +--------------+--------------------------------------------------------------+
  26 |test method   | This test case kills the processes of load balance service   |
  27 |              | on a selected control node, then checks whether the request  |
  28 |              | of the related Openstack command is OK and the killed        |
  29 |              | processes are recovered.                                     |
  30 +--------------+--------------------------------------------------------------+
  31 |attackers     | In this test case, an attacker called "kill-process" is      |
  32 |              | needed. This attacker includes three parameters:             |
  33 |              | 1) fault_type: which is used for finding the attacker's      |
  34 |              | scripts. It should be always set to "kill-process" in this   |
  35 |              | test case.                                                   |
  36 |              | 2) process_name: which is the process name of the specified  |
  37 |              | OpenStack service. If there are multiple processes use the   |
  38 |              | same name on the host, all of them are killed by this        |
  39 |              | attacker.                                                    |
  40 |              | In this case. This parameter should always set to "swift-    |
  41 |              | proxy".                                                      |
  42 |              | 3) host: which is the name of a control node being attacked. |
  43 |              |                                                              |
  44 |              | e.g.                                                         |
  45 |              | -fault_type: "kill-process"                                  |
  46 |              | -process_name: "haproxy"                                     |
  47 |              | -host: node1                                                 |
  48 |              |                                                              |
  49 +--------------+--------------------------------------------------------------+
  50 |monitors      | In this test case, two kinds of monitor are needed:          |
  51 |              | 1. the "openstack-cmd" monitor constantly request a specific |
  52 |              | Openstack command, which needs two parameters:               |
  53 |              | 1) monitor_type: which is used for finding the monitor class |
  54 |              | and related scritps. It should be always set to              |
  55 |              | "openstack-cmd" for this monitor.                            |
  56 |              | 2) command_name: which is the command name used for request. |
  57 |              |                                                              |
  58 |              | 2. the "process" monitor check whether a process is running  |
  59 |              | on a specific node, which needs three parameters:            |
  60 |              | 1) monitor_type: which used for finding the monitor class    |
  61 |              | and related scripts. It should be always set to "process"    |
  62 |              | for this monitor.                                            |
  63 |              | 2) process_name: which is the process name for monitor       |
  64 |              | 3) host: which is the name of the node runing the process    |
  65 |              | In this case, the command_name of monitor1 should be         |
  66 |              | services that is supported by load balancer and the process- |
  67 |              | name of monitor2 should be "haproxy", for example:           |
  68 |              |                                                              |
  69 |              | e.g.                                                         |
  70 |              | monitor1:                                                    |
  71 |              | -monitor_type: "openstack-cmd"                               |
  72 |              | -command_name: "nova image-list"                             |
  73 |              | monitor2:                                                    |
  74 |              | -monitor_type: "process"                                     |
  75 |              | -process_name: "haproxy"                                     |
  76 |              | -host: node1                                                 |
  77 |              |                                                              |
  78 +--------------+--------------------------------------------------------------+
  79 |metrics       | In this test case, there are two metrics:                    |
  80 |              | 1)service_outage_time: which indicates the maximum outage    |
  81 |              | time (seconds) of the specified Openstack command request.   |
  82 |              | 2)process_recover_time: which indicates the maximun time     |
  83 |              | (seconds) from the process being killed to recovered         |
  84 |              |                                                              |
  85 +--------------+--------------------------------------------------------------+
  86 |test tool     | Developed by the project. Please see folder:                 |
  87 |              | "yardstick/benchmark/scenarios/availability/ha_tools"        |
  88 |              |                                                              |
  89 +--------------+--------------------------------------------------------------+
  90 |references    | ETSI NFV REL001                                              |
  91 |              |                                                              |
  92 +--------------+--------------------------------------------------------------+
  93 |configuration | This test case needs two configuration files:                |
  94 |              | 1) test case file: opnfv_yardstick_tc053.yaml                |
  95 |              | -Attackers: see above "attackers" discription                |
  96 |              | -waiting_time: which is the time (seconds) from the process  |
  97 |              | being killed to stoping monitors the monitors                |
  98 |              | -Monitors: see above "monitors" discription                  |
  99 |              | -SLA: see above "metrics" discription                        |
 100 |              |                                                              |
 101 |              | 2)POD file: pod.yaml                                         |
 102 |              | The POD configuration should record on pod.yaml first.       |
 103 |              | the "host" item in this test case will use the node name in  |
 104 |              | the pod.yaml.                                                |
 105 |              |                                                              |
 106 +--------------+--------------------------------------------------------------+
 107 |test sequence | description and expected result                              |
 108 |              |                                                              |
 109 +--------------+--------------------------------------------------------------+
 110 |step 1        | start monitors:                                              |
 111 |              | each monitor will run with independently process             |
 112 |              |                                                              |
 113 |              | Result: The monitor info will be collected.                  |
 114 |              |                                                              |
 115 +--------------+--------------------------------------------------------------+
 116 |step 2        | do attacker: connect the host through SSH, and then execute  |
 117 |              | the kill process script with param value specified by        |
 118 |              | "process_name"                                               |
 119 |              |                                                              |
 120 |              | Result: Process will be killed.                              |
 121 |              |                                                              |
 122 +--------------+--------------------------------------------------------------+
 123 |step 3        | stop monitors after a period of time specified by            |
 124 |              | "waiting_time"                                               |
 125 |              |                                                              |
 126 |              | Result: The monitor info will be aggregated.                 |
 127 |              |                                                              |
 128 +--------------+--------------------------------------------------------------+
 129 |step 4        | verify the SLA                                               |
 130 |              |                                                              |
 131 |              | Result: The test case is passed or not.                      |
 132 |              |                                                              |
 133 +--------------+--------------------------------------------------------------+
 134 |post-action   | It is the action when the test cases exist. It will check    |
 135 |              | the status of the specified process on the host, and restart |
 136 |              | the process if it is not running for next test cases.        |
 137 |              |                                                              |
 138 |              | Notice: This post-action uses 'lsb_release' command to check |
 139 |              | the host linux distribution and determine the OpenStack      |
 140 |              | service name to restart the process. Lack of 'lsb_release'   |
 141 |              | on the host may cause failure to restart the process.        |
 142 |              |                                                              |
 143 +--------------+--------------------------------------------------------------+
 144 |test verdict  | Fails only if SLA is not passed, or if there is a test case  |
 145 |              | execution problem.                                           |
 146 |              |                                                              |
 147 +--------------+--------------------------------------------------------------+