EVPN-VXLAN Lightweight Leaf to Server Loop Detection

How Lightweight PE-CE Loop Detect Works

When you configure lightweight PE-CE loop detect, the device transmits periodic multicast protocol data units (PDUs) on PE to CE interfaces for detecting loops. The device can then block the interface upon receiving these self-generated PDUs. When the device receives a loop detect PDU, it breaks the loop by blocking (operationally shutting down) the ingress port.

The loop detect PDUs use the Connectivity Fault Management (CFM) protocol PDU format, although you don't explicitly configure CFM with this feature. The loop detect error messages logged by this feature include the CFM keyword, such as CFMD_LOOP_DETECTED and CFMD_LOOP_CLEARED.

We recommend that you enable lightweight PE-CE loop detect initially before you configure EVPN-VXLAN, so you can detect any loops and take corrective actions before EVPN traffic is flowing through the network. When this feature detects loops, the device raises loop detect error messages immediately. However, if you bring up a large-scale EVPN network that already contains loops, even with this feature enabled, the interface doesn't come down immediately and traffic continues to flow through the loop for some time while the network stabilizes.

If a loop is introduced into the network later in a stable running EVPN-VXLAN fabric, this feature will detect the loop and stop traffic flow through the loop immediately.

Options for Repairing Loops

If required, break and clear the loop. To bring the interface back online, you can configure a revert interval using the revert-interval seconds statement at the [edit protocols loop-detect interface name] hierarchy level. When the revert interval expires, the device automatically brings the interface back online. The default revert interval is 0 seconds, which means the interval never expires and the interface doesn't automatically revert to its prior state.

If you don’t explicitly configure a revert interval other than 0, the port never reverts to its state before the loop detect event and action. To manually bring the interface back online, you must clear the status using the clear loop-detect enhanced interface name command.

Supported Interface Configurations

To configure this lightweight PE-CE loop detect feature, specify a logical interface name. We don't support this feature with physical interfaces, only with logical interfaces as follows:

• Enterprise style interface configurations without flexible Ethernet services—Only with logical unit 0.

• Enterprise style interface configurations with flexible Ethernet services—On any logical interfaces you can configure on the device, including logical interfaces for a trunk interface with the native VLAN ID and other configured VLANs.

• Service provider style interface configurations—Only on QFX10002-60C, QFX10002, QFX10008, and QFX10016 switches, starting in Junos OS Release 22.4R1. We don't support this lightweight loop detect feature on service provider style interfaces with any other devices.

• Aggregated Ethernet interfaces—On a logical unit X of an aggregated Ethernet interface (aeN.X). With enterprise style aggregated Ethernet interface configurations without flexible Ethernet services configured, we only support logical unit 0. Otherwise you can use any configured logical unit X.

  On aggregated Ethernet interfaces with Link Aggregation Control Protocol (LACP) enabled, the LACP state remains up (Collecting or Distributing) even if the loop detect action brings the logical interface down.

See Flexible Ethernet Services Encapsulation for more on flexible Ethernet services, enterprise style interface configurations, and service provider style interface configurations.

Lightweight PE-CE Loop Detect Scenarios

The following three lightweight PE-CE loop detect scenarios demonstrate that loops can form with different Ethernet segment identifiers (ESIs), with the same ESI, or with no ESI.

• Different ESI Looped
• Same ESI Looped
• No ESI on Looped Ports

Figure 1: Different ESI Looped

Figure 2: Same ESI Looped

Figure 3: No ESI on Looped Ports

Lightweight PE-CE Loop Detect Use Cases using Layer 2 Heartbeats

In this section we show the following use cases with lightweight PE-CE loop detect enabled:

• Use Case 1: The loop is occurring through the switch due to misconfiguration of the switch.

• Use Case 2: The loop is caused by misalignment of cable connections on the switch.

In both use cases, the functionality is not dependent on the BGP speed of control-plane advertisement, and the lightweight PE-CE loop detect is independent of the configured ESI values.

• EVPN-VXLAN Lightweight PE-CE Loop Detect Use Case 1
• EVPN-VXLAN Lightweight PE-CE Loop Detect Use Case 2

Figure 4: Scenario 1

Figure 5: Scenario 2

Enable Lightweight PE-CE Loop Detect on a Logical Interface

To enable lightewight loop detect for a logical interface or for all logical interfaces, use the loop-detect statement at the [edit protocols] hierarchy level. Include a supported loop-detect-action for the interface(s) and optionally specify a vlan-id, as follows:

You can also optionally set the following values at the [edit protocols loop-detect enhanced interface (logical-interface-name | all)] hierarchy level:

• The revert-interval option—After you repair the loop, the device brings the interface(s) up again after this interval expires (default is 0 seconds).

• The transmit-interval option—Customize how often to transmit loop detect PDUs (default is 1 second, see loop-detect for more on the values you can set for this interval).

Starting with Junos 24.4 we’ve added support for monitoring all VLANs on a logical interface with the vlan-id all configuration statement option at the [edit protocols loop-detect enhanced interface] hierarchy level. This enhancement detects network loops across multiple VLANs and interfaces, improving network stability and performance in scaled environments.

When you configure the vlan-id all option on supported devices, the devices show the following behavioral changes for this feature:

• The revert-interval configuration is not effective for scaled loop-detect sessions, which makes them non-revertive. You must issue the clear loop-detect enhanced interface command to clear the loop condition.

• The receive statistics for loop-detect PDUs do not increment for scaled loop-detect sessions during a loop condition.

• We support only a 1-second transmit interval for scaled loop-detect sessions.

See Feature Explorer for the platforms that support lightweight PE-CE loop detect as follows:

• Lightweight PE-CE Loop Detection on EVPN-VXLAN Fabrics—Support for lightweight PE-CE loop detect

• Lightweight Loop Detection with Scale—Support for scaled lightweight PE-CE loop detect with the vlan-id all option

Sample Configuration with Trunk Mode Enterprise Style Interface

The following sample configuration enables lightweight PE-CE loop detect on interface ge-0/0/1.0, which is a trunk interface with vlan-id 100.

Sample Configuration with Service Provider Style Interface

The following sample configuration enables loop detect for vlan-id 100 on provider edge (PE) and CE devices with service provider style interface configurations. This configuration doesn't specify a revert interval. As a result, after the device detects a loop and you correct the loop, enter the clear loop detect enhanced interface name command to bring the interface back online.

PE device configuration:

CE device configuration:

CLI Commands to Display or Clear Loop Detect Status

Use the show loop-detect enhanced interface command to display loop status on an interface or all interfaces.

Use the clear loop-detect enhanced interface command to restore an interface or all interfaces to their prior state after the device detects a loop and applies a configured action to break the loop.

• Show command without any loop
• Show command with loop detect status