NSX Guest Introspection capacity and troubleshooting

Products

• VMware NSX

Issue/Introduction

Symptoms:

• GI SVM memory exceeds 75% and guest workloads may be impacted.
• Log messages on the GI SVM show that connection rates exceed capacity.
• UI alerts will show if the GI SVM has exceeded memory threshold.
• In the vsm.log file, you see entries similar to:

2017-12-28 05:08:07.203 GMT  INFO messagingTaskExecutor-1 EventServiceImpl:128 - [SystemEvent] Time:'Thu Dec 28 05:08:07.200 GMT 2017', Severity:'Critical', Event Source:'Guest Introspection Health Status', Code:'260009', Event Message:'High memory usage on GI SVM on host-3188.', Module:'Guest Introspection Health Status', Universal Object:'false'

• In the Networking & Security > Events > System Events, you see a recorded event of Code:'260009' and Event Message:'High memory usage on GI SVM on host-XXXX.'

Note: Use the vSphere Managed Object Browser to identify the problem host from the host moid (host-XXXX) logged in the Event Message.

Environment

VMware NSX for vSphere 6.3.x
VMware NSX for vSphere 6.4.x

Cause

This issue occurs when deployed virtual machines / host ratio or network connections per virtual machine capacity has been exceeded

Resolution

To resolve this issue:

1. Identify total throughput of network events on the GI SVM and specific VMs generating high rate of connections.

• Collect the GI SVM logs. For more information, see How to collect logs in NSX for vSphere 6.x Guest Introspection
• On a 10m interval, the GI SVM logs total processed network connections. You see entries similar to: 
   2017-12-28 05:28:06.934 UTC  INFO taskScheduler-3 EventStatsManager:185 - Event counts
  Total: 11820217
  Preconnect: 5913188
  Connect: 0
  Listenstart: 0
  Listenstop: 0
  Inboundconnect: 0
  Disconnect: 0
• Determine the rate of network events by observing the change in "Total" count of events.
• To determine high-throughput VMs, enable DEBUG logs on the GI SVM. For more information, see How to collect logs in NSX for vSphere 6.x Guest Introspection
• On a 10m interval, the GI SVM logs per VM, the total processed network connections. You see entries similar to: 
  2017-12-28 05:28:06.934 UTC  INFO taskScheduler-3 EventStatsManager:198 - Per VM individual event counts
  VM UUID, VMX Path: <VM UUID, VM_vmx_path>
  Total: 11820217
  ......

2. Identify hosts where virtual machines/ host has exceeded capacity.
3. Redistribute virtual machines across hosts for load balancing.

This table provides the capacity performance limits for GI SVM.
 

GI Memory Configuration	VMs per ESXi	Sustained Network Events Rate per VM (events/s)	AD Groups per User
2 GB	50	1850	4
2 GB	50	1000	4
2 GB	50	800	4
2 GB	50	700	4

Note: This is applicable for NSX for vSphere 6.3.5 and above versions only.

Additional Information

It is possible to adjust the CPU/Memory thresholds on the GI SVM.  You will see logs and system events once the threshold values are exceeded.

Get current global threshold:
Method : GET
URL: https://NSXM_IP/api/2.0/endpointsecurity/usvmstats/usvmhealththresholds

Set new threshholds:
Method: PUT
URL: https://NSXM_IP/api/2.0/endpointsecurity/usvmstats/usvmhealththresholds
Body:
<?xml version="1.0" encoding="UTF-8"?>
<UsvmHealthStats>
    <memThreshold>75</memThreshold>
    <cpuThreshold>75</cpuThreshold>
</UsvmHealthStats>