Connection failed dropped by load balancer when accessing app hosted on server pool behind NSX load balancer
Article ID: 411211
Updated On:
Products
Issue/Introduction
- Experience connection timeouts when trying to connect to an NSX Native Load Balancer Virtual Server created by TKGI
- Traffic analysis in NSX UI shows drops caused by an NSX rule when testing connectivity between pool members and VIP
- Load balancer debugging shows no active connections or sessions in the session table
- Small load balancer configured with more than 20 virtual servers
Steps to validate:
- Navigate to Plan & Troubleshoot > Traffic Analysis > Traceflow in NSX UI
- Test traffic by pinging between pool member and VIP - observe drops
- Check the load balancer session table using:
get load-balancer <lb_uuid> session-tables | count l4lb - Verify the number of virtual servers configured on the load balancer
Environment
VMware NSX
VMware Tanzu Kubernetes Grid
Cause
The load balancer exceeds VMware NSX ConfigMax limits. Small load balancers configured with more than 20 virtual servers exceed the supported configuration, causing connection failures and traffic drops. Load balancer capacity is also limited by the Edge Node form factor hosting the load balancer instances.
Resolution
Reduce the number of virtual servers to comply with VMware NSX ConfigMax limits or upgrade the load balancer size:
Option 1: Reduce Virtual Servers
- Configure 20 or fewer virtual servers on small load balancers per VMware NSX ConfigMax guidelines
Option 2: Upgrade Load Balancer Size
- Upgrade to medium load balancer (supports up to 100 virtual servers)
- Upgrade to large load balancer (supports up to 1,000 virtual servers)
- Upgrade to extra large load balancer (supports up to 2,000 virtual servers)
Option 3: Consider Edge Node Form Factor Limits
- Verify the Edge Node form factor can support the required number of load balancer instances
- Small Edge Node (VM): 1 small load balancer instance maximum
- Medium Edge Node (VM): 10 small or one medium load balancer instance maximum
- Large Edge Node (VM): 40 small, four medium, or one large load balancer instance maximum
- Extra Large Edge Node (VM): 80 small, eight medium, two large, or one extra large load balancer instance maximum
- Bare-Metal Edge Node: 750 small, 75 medium, 18 large, or nine extra large load balancer instances maximum
ConfigMax Limits per Load Balancer Size:
- Small: 20 virtual servers
- Medium: 100 virtual servers
- Large: 1,000 virtual servers
- Extra Large: 2,000 virtual servers
Important: Configurations exceeding these limits may work intermittently but are unsupported. Operating within ConfigMax limits ensures reliable operation and continued support.
If the error persists after following these steps, contact Broadcom Support for further assistance.
Additional Information
Feedback
