This section discusses the steps to troubleshoot non-uniform traffic distribution between two GSLB sites when the round-robin method of GSLB load balancing is used.

  1. Non-uniform RR Traffic Distribution could be due to other reasons as discussed in Non-uniform RR Traffic Distribution in the VMware NSX Advanced Load Balancer Monitoring and Operability Guide. Identify the potential causes and validate if the load balancing behavior is correct.

  2. After following the guidelines mentioned in step 1, if the problem is still observed, get the output of show virtualservice as follows:

    show virtualservice <for each DNS VS  this GS subdomain is delegated to > gslbserviceinternal filter gs_ref <gs-name> se_ref <for each SE this DNS VS is placed> disable_aggregate core

    1. Make a note of total_connections for DNS VS across two sites and calculate the ratio for C0 and C1 accordingly. For an example (for core C0):

      +---------------------------------------+-------------------------------------+
| Field                                 | Value                               |
+---------------------------------------+-------------------------------------+
<...Truncated...>                                                             |
| server_internal[1]                    |                                     |
|   ip_addr                             | 100.64.218.68                       |
|   port                                | 0                                   |
|   oper_status                         |                                     |
|     state                             | OPER_UP                             |
|   ref_count                           | 6                                   |
|   server_stats                        |                                     |
|     current_connections               | 0                                   |
|     total_connections                 | 0                                   |
|     pre_closed_connections            | 0                                   |
|     bad_connections                   | 0                                   |
|     inlhm_skipped                     | 0                                   |
|     skipped_count                     | 0                                   |
|     response_time                     | 0                                   |
|     response_time_var                 | 0                                   |
|     inlhm_down_duration               | 0                                   |
|     inlhm_good_ewma                   | 0                                   |
|     inlhm_bad_ewma                    | 0                                   |
|     no_lport                          | 0                                   |
|     inlhm_letthrough                  | 0                                   |
|     current_load                      | 0                                   |
|     inlhm_skip_bypassed               | 0                                   |
|     current_tasks                     | 0                                   |
|     current_tasks_fb                  | 2147483647                          |
|     rl_good_fb                        | 0                                   |
|     rl_bad_fb                         | 0                                   |
|     capest_rand_max_conn_skipped      | 0                                   |
|     skipped_get_next_count            | 0                                   |
|   proc_id                             | C0                                  |
|   se_uuid                             | Avi-Service-Engine:se-0050568148ce  |
| server_internal[2]                    |                                     |
|   ip_addr                             | 100.64.218.64                       |
|   port                                | 0                                   |
|   oper_status                         |                                     |
|     state                             | OPER_UP                             |
|   ref_count                           | 6                                   |
|   server_stats                        |                                     |
|     current_connections               | 0                                   |
|     total_connections                 | 0                                   |
|     pre_closed_connections            | 0                                   |
|     inlhm_skipped                     | 0                                   |
|     skipped_count                     | 0                                   |
|     response_time                     | 0                                   |
|     response_time_var                 | 0                                   |
|     inlhm_down_duration               | 0                                   |
|     inlhm_good_ewma                   | 0                                   |
|     inlhm_bad_ewma                    | 0                                   |
|     no_lport                          | 0                                   |
|     inlhm_letthrough                  | 0                                   |
|     current_load                      | 0                                   |
|     inlhm_skip_bypassed               | 0                                   |
|     current_tasks                     | 0                                   |
|     current_tasks_fb                  | 2147483647                          |
|     rl_good_fb                        | 0                                   |
|     rl_bad_fb                         | 0                                   |
|     capest_rand_max_conn_skipped      | 0                                   |
|     skipped_get_next_count            | 0                                   |
|   proc_id                             | C0                                  |
|   se_uuid                             | Avi-Service-Engine:se-0050568148ce  |
+---------------------------------------+-------------------------------------+

      Ratio is calculated as 34503766/35105210 = 0.98

    • In this case, skew of 2 percent over a long period of time can be attributed to member state flaps. If one member is down, the other member will take the load.