docs/yardstick/opnfv_yardstick_tc019.rst

   1 *************************************
   2 Yardstick Test Case Description TC019
   3 *************************************
   4
   5 +-----------------------------------------------------------------------------+
   6 |Control Node Openstack Service High Availability                             |
   7 |                                                                             |
   8 +--------------+--------------------------------------------------------------+
   9 |test case id  | OPNFV_YARDSTICK_TC019_HA: Control node Openstack service down|
  10 |              |                                                              |
  11 +--------------+--------------------------------------------------------------+
  12 |test purpose  | This test case will verify the high availability of the      |
  13 |              | service provided by OpenStack (like nova-api, neutro-server) |
  14 |              | on control node.                                             |
  15 |              |                                                              |
  16 +--------------+--------------------------------------------------------------+
  17 |test method   | This test case kills the processes of a specific Openstack   |
  18 |              | service on a selected control node, then checks whether the  |
  19 |              | request of the related Openstack command is OK and the killed|
  20 |              | processes are recovered.                                     |
  21 |              |                                                              |
  22 +--------------+--------------------------------------------------------------+
  23 |attackers     | In this test case, an attacker called "kill-process" is      |
  24 |              | needed. This attacker includes three parameters:             |
  25 |              | 1) fault_type: which is used for finding the attacker's      |
  26 |              | scripts. It should be always set to "kill-process" in this   |
  27 |              | test case.                                                   |
  28 |              | 2) process_name: which is the process name of the specified  |
  29 |              | OpenStack service. If there are multiple processes use the   |
  30 |              | same name on the host, all of them are killed by this        |
  31 |              | attacker.                                                    |
  32 |              | 3) host: which is the name of a control node being attacked. |
  33 |              |                                                              |
  34 |              | e.g.                                                         |
  35 |              | -fault_type: "kill-process"                                  |
  36 |              | -process_name: "nova-api"                                    |
  37 |              | -host: node1                                                 |
  38 |              |                                                              |
  39 +--------------+--------------------------------------------------------------+
  40 |monitors      | In this test case, two kinds of monitor are needed:          |
  41 |              | 1. the "openstack-cmd" monitor constantly request a specific |
  42 |              |    Openstack command, which needs two parameters:            |
  43 |              | 1) monitor_type: which is used for finding the monitor class |
  44 |              | and related scritps. It should be always set to              |
  45 |              | "openstack-cmd" for this monitor.                            |
  46 |              | 2) command_name: which is the command name used for request  |
  47 |              |                                                              |
  48 |              | 2. the "process" monitor check whether a process is running  |
  49 |              |    on a specific node, which needs three parameters:         |
  50 |              | 1) monitor_type: which used for finding the monitor class and|
  51 |              | related scritps. It should be always set to "process"        |
  52 |              | for this monitor.                                            |
  53 |              | 2) process_name: which is the process name for monitor       |
  54 |              | 3) host: which is the name of the node runing the process    |
  55 |              |                                                              |
  56 |              | e.g.                                                         |
  57 |              | monitor1:                                                    |
  58 |              | -monitor_type: "openstack-cmd"                               |
  59 |              | -command_name: "nova image-list"                             |
  60 |              | monitor2:                                                    |
  61 |              | -monitor_type: "process"                                     |
  62 |              | -process_name: "nova-api"                                    |
  63 |              | -host: node1                                                 |
  64 |              |                                                              |
  65 +--------------+--------------------------------------------------------------+
  66 |metrics       | In this test case, there are two metrics:                    |
  67 |              | 1)service_outage_time: which indicates the maximum outage    |
  68 |              | time (seconds) of the specified Openstack command request.   |
  69 |              | 2)process_recover_time: which indicates the maximun time     |
  70 |              | (seconds) from the process being killed to recovered         |
  71 |              |                                                              |
  72 +--------------+--------------------------------------------------------------+
  73 |test tool     | Developed by the project. Please see folder:                 |
  74 |              | "yardstick/benchmark/scenarios/availability/ha_tools"        |
  75 |              |                                                              |
  76 +--------------+--------------------------------------------------------------+
  77 |references    | ETSI NFV REL001                                              |
  78 |              |                                                              |
  79 +--------------+--------------------------------------------------------------+
  80 |configuration | This test case needs two configuration files:                |
  81 |              | 1) test case file: opnfv_yardstick_tc019.yaml                |
  82 |              | -Attackers: see above "attackers" discription                |
  83 |              | -waiting_time: which is the time (seconds) from the process  |
  84 |              | being killed to stoping monitors the monitors                |
  85 |              | -Monitors: see above "monitors" discription                  |
  86 |              | -SLA: see above "metrics" discription                        |
  87 |              |                                                              |
  88 |              | 2)POD file: pod.yaml                                         |
  89 |              | The POD configuration should record on pod.yaml first.       |
  90 |              | the "host" item in this test case will use the node name in  |
  91 |              | the pod.yaml.                                                |
  92 |              |                                                              |
  93 +--------------+--------------------------------------------------------------+
  94 |test sequence | description and expected result                              |
  95 |              |                                                              |
  96 +--------------+--------------------------------------------------------------+
  97 |step 1        | start monitors:                                              |
  98 |              | each monitor will run with independently process             |
  99 |              |                                                              |
 100 |              | Result: The monitor info will be collected.                  |
 101 |              |                                                              |
 102 +--------------+--------------------------------------------------------------+
 103 |step 2        | do attacker: connect the host through SSH, and then execute  |
 104 |              | the kill process script with param value specified by        |
 105 |              | "process_name"                                               |
 106 |              |                                                              |
 107 |              | Result: Process will be killed.                              |
 108 |              |                                                              |
 109 +--------------+--------------------------------------------------------------+
 110 |step 3        | stop monitors after a period of time specified by            |
 111 |              | "waiting_time"                                               |
 112 |              |                                                              |
 113 |              | Result: The monitor info will be aggregated.                 |
 114 |              |                                                              |
 115 +--------------+--------------------------------------------------------------+
 116 |step 4        | verify the SLA                                               |
 117 |              |                                                              |
 118 |              | Result: The test case is passed or not.                      |
 119 |              |                                                              |
 120 +--------------+--------------------------------------------------------------+
 121 |post-action   | It is the action when the test cases exist. It will check the|
 122 |              | status of the specified process on the host, and restart the |
 123 |              | process if it is not running for next test cases             |
 124 |              |                                                              |
 125 +--------------+--------------------------------------------------------------+
 126 |test verdict  | Fails only if SLA is not passed, or if there is a test case  |
 127 |              | execution problem.                                           |
 128 |              |                                                              |
 129 +--------------+--------------------------------------------------------------+