Within NSX, intermittent drops on north-south traffic

Products

• VMware NSX

Issue/Introduction

• A Gateway QoS profile is configured.
• Affected traffic is traversing a T1 where QoS profile is applied.
• The T1 uplink interface shows active drops on tx_drop_rate_limit.

get logical-router interface <uuid> stats | json
...
"rx_drop_kni": 0,
"rx_drop_redirect": 0,
"rx_drop_rate_limit": 10,
"rx_frags": 0,
"rx_frag_ok": 0,
"rx_frag_error": 0,
"rx_frag_timeout": 0,
"rx_drop_ipsec": 0,
"rx_drop_ipsec_nosa": 0,
"rx_drop_ipsec_novti": 0,
"rx_rs": 0,
"rx_mld": 0,
"rx_drop_igmp": 0,
"tx_drop_rate_limit": 1005761, <<<<<<<<<<<<<<<
...

• Removing QoS profile from the T1 resolves the connectivity issue.

 

Environment

VMware NSX-T Data Center
VMware NSX

Resolution

Based on the QoS configuration (committed rate, burst size etc) this behaviour may be expected.

Additional Information

If an issue with QoS is suspected, the following data will be required for further analysis.

• iperf output showing the throughput from source to destination.
• Details of the problematic QoS profile. 
• Details of all other QoS profiles within the environment and T1s on which they're attached
• Multiple samples of the following commands (must be run on Edge which contains the active T1)
  • get dataplane cpu stats 
  • get dataplane throughput 10
  • get physical-port <fp-ethX> stats
  • get logical-router interface <lrport-uuid> stats (Logical Router uplink where QoS is applied)
• Support bundle for the NSX Managers, Edges and hosts on which the Edges are running.