Did the Network Layer Drop My Packets?

When a request silently disappears, these panels tell you in seconds whether the network is responsible — and if so, which layer.

IP panels vs vNIC panels

ZEC exposes two different views of an instance's network traffic:

• IP panels measure north-south traffic — bytes and packets crossing in or out of the public IP attached to your instance. They count only traffic that enters or leaves the ZEC network through that IP, and the rate-limit policer that enforces your IP bandwidth quota lives at this layer.

• vNIC panels measure everything the virtual NIC sees, from the NIC's perspective. That includes north-south traffic and east-west traffic to other instances, services, and private endpoints inside ZEC — VPC traffic, intra-region service calls, and so on. The IP panels never see this traffic.

A practical rule:

• Drops on IP panels only → almost certainly your IP bandwidth quota. Raise the IP bandwidth or smooth the burst.

• Drops on vNIC panels only → the affected traffic is internal and never touched the IP policer. Cross-check with the relevant private network panels and contact support if drops persist.

• Drops on both → the traffic was north-south and hit the IP limit; the vNIC panels are echoing the same loss.

IP Bandwidth — Inbound / Outbound drop bandwidth

IP Packet Transmission — Inbound / Outbound packet loss

Traffic that hit the IP and was dropped by the data plane. Drops here are almost always bandwidth-limit drops: your traffic exceeded the IP's contracted rate and the policer dropped the burst.

This is the panel to check first when you see request loss on a workload that runs in bursts (TCP slow-start, sudden fan-in, batch jobs). If these panels show drops, raise the IP bandwidth or smooth the traffic. If they are clean, the problem is somewhere else.

IP Bandwidth — Inbound / Outbound

IP Packet Transmission — Inbound / Outbound

Total north-south throughput on the IP. Use as the denominator for the drop ratio above and as a headline view of how much traffic the instance is moving across its public boundary.

vNIC Bandwidth — Inbound / Outbound

vNIC Packet Transmission — Inbound / Outbound

The full traffic the virtual NIC handled — north-south and east-west combined. Direction is normalized to the guest's perspective: Inbound is what the VM received.

If a vNIC bandwidth or packet number is much higher than the corresponding IP number, the difference is east-west traffic. This is normal for instances that talk to other instances in the same VPC.

vNIC Packet Transmission — Inbound / Outbound packet loss

vNIC Packet Transmission — Inbound error

Drops or errors at the virtual NIC level. These should be zero in normal operation. If you see non-zero values, contact support — they are infrastructure-side red flags.

API Reference

IP metrics use DescribeEipMonitorData. vNIC metrics use DescribeInstanceMonitorData. The EIP API returns both throughput and drop values (in, out, loseIn, loseOut) in a single response.

Console Panel	API	metricType	Notes
IP Bandwidth (Inbound / Outbound / Drops)	DescribeEipMonitorData	BANDWIDTH	Response includes in, out, loseIn, loseOut
IP Packet Transmission (Inbound / Outbound / Drops)	DescribeEipMonitorData	PACKETS	Response includes in, out, loseIn, loseOut
vNIC Bandwidth — Inbound / Outbound	—	—	Console only
vNIC Packet Transmission — Inbound	DescribeInstanceMonitorData	RX_PACKETS
vNIC Packet Transmission — Outbound	DescribeInstanceMonitorData	TX_PACKETS
vNIC Packet Transmission — Inbound packet loss	DescribeInstanceMonitorData	RX_DROPPED
vNIC Packet Transmission — Outbound packet loss	DescribeInstanceMonitorData	TX_DROPPED
vNIC Packet Transmission — Inbound error	DescribeInstanceMonitorData	RX_ERRORS