EMAIL SUPPORT

dclessons@dclessons.com

LOCATION

US

PBR Traffic Flow Learning

PBR Traffic Flow Learning

Depending on the PBR design, several traffic flows are possible. These figures depict the east-west traffic flow between endpoints in EPG client and EPG web, when a firewall service node in Layer 3 mode is integrated in Cisco ACI fabric using a service graph with PBR.

EPG client is a consumer EPG and the EPG web is a provider EPG, using a contract with service graph PBR. The generated traffic from the client endpoint that is destined for the web endpoint follows this sequence:

  1. Endpoint client, which is connected to Leaf 1, sends traffic to the web.

  2. Assuming that Leaf 1 has not learned the destination endpoint, Leaf 1 forwards the traffic to the spine proxy. Currently, Leaf1 cannot resolve the destination EPG class ID (pcTag) either to apply a contract with PBR policy.

  3. The spine node forwards the traffic to Leaf 3, to which the destination web endpoint is connected. In addition, Leaf 3 learns the source endpoint (client) information from this traffic and populates its endpoint table.

  4. Since Leaf 3 can resolve both, the source and destination EPG class IDs, the contract with the PBR service graph is invoked, which results with PBR traffic redirection on Leaf 3. In this process, the destination MAC address is rewritten to the PBR node MAC address (MAC Ext of the firewall external interface) on the consumer side. Leaf 3 looks up the PBR node MAC Ext in the PBR BD. Since Leaf 3 does not know where this destination MAC address is connected, the traffic goes to the spine proxy, which forwards the traffic to Leaf 2, to which the PBR node is connected. Hence, the traffic is forwarded to the external interface of the PBR node for inspection and routing. Leaf 2 does not learn the client IP address from this traffic like Leaf 3 did, because IP data-plane learning is disabled for the PBR node bridge domain.

The traffic from the PBR node that is destined for the web endpoint follows this sequence:

GENERAL FAQ

PBR is used in an ACI service chain to steer the traffic to a particular service node (eg, firewall) based on contract policies, even if other normal routing or endpoint learning does not send the traffic through that device.

If the destination endpoint is not known, traffic is first sent from the ingress leaf to the spine proxy. The spine forwards it to the leaf where the web endpoint is running. After the leaf sends both EPG class IDs, the PBR policy redirects the traffic to the firewall service node.

The leaf changes the destination MAC to that MAC EXT of the firewall external interface. This allows the traffic to be sent to the PBR node (firewall) for examination and processing, applying the service graph policy.

IP data-plane learning is intentionally disabled on the PBR bridge domain. This avoids populating endpoint tables with PBR node traffic, which could cause incorrect forwarding decisions.

Because the PBR node’s MAC is not present in the leaf’s endpoint table, the packet goes to the spine proxy, which forwards it to the correct leaf where the PBR node (firewall) is connected.


Comment

    You are will be the first.

LEAVE A COMMENT

Please login here to comment.