VS takes long time to function after disable/enable in NSX-T Cloud

Products

• VMware Avi Load Balancer

Issue/Introduction

Symptom:

In a scaled NSX-T Cloud environment, it may be observed that the VS may take a few minutes to be marked OPER_UP after a disable/enable.

In the VS Events, it can be seen that the CC_DETACH_IP_SUCCESS takes several minutes since the VS was disabled. Same behavior may be seen for CC_ATTACH_IP_SUCCESS Event. (See example below)

Environment

NSX-T Cloud in 22.1.x versions 

Cause

This is an expected outcome in a scaled environment. The reconcile task takes several minutes blocking other tasks from running (Attach/Detach IP tasks). This is because AVI creates the following NS Service and NS Groups in the NSX-T. This may consume significant amount of time to finish up delaying the Attach/Detach IP tasks.

• NSService Objects

1. One for the Controller cluster
2. One per VS
3. One per Pool

• NSGroup Objects

1. One for the Controller Cluster
2. One per VS
3. One per VS's SEs
4. One for All SE Mgmt IPs
5. One for All SE IPs (data vNic + Mgmt vNic)
6. One per SE group

 

Resolution

In 30.2.2, a flag called "automate_dfw_objects" is introduced which can be turned off to disable the creation of the above objects by AVI. This can significantly reduce the reconcile thereby leading to faster IP Attach/Detach.

If the DFW objects are needed, then please follow the recommended configuration for reconcile and discovery as per the following documents.

> configure cloud <cloud-name>
> autoscale_polling_interval 600
> nsxt_configuration
> no automate_dfw_objects
> save
> save



> configure controller properties
> cloud_reconcile_interval 1440
> cloud_discovery_interval 30
> save

 

Doc: https://docs.vmware.com/en/VMware-Avi-Load-Balancer/30.2/Installation-Guide/GUID-C62EAFC3-7966-4B8B-AA44-F38F428DAA1E.html#one-arm-mode-with-overlay-vip-segment--single-tier-1-1