Uploaded image for project: 'OpenShift Request For Enhancement'
  1. OpenShift Request For Enhancement
  2. RFE-4951

EgressIP support for single pod egressing multiple networks


    • Icon: Feature Request Feature Request
    • Resolution: Done
    • Icon: Critical Critical
    • openshift-4.17
    • openshift-4.14
    • SDN
    • None
    • False
    • None
    • False
    • Not Selected

      1. Proposed title of this feature request

      EgressIP support for single pod egressing multiple networks

      2. What is the nature and description of the request?

      The context of this RFE is baremetal clusters with a relatively large number of nodes (can be 100+). Networking includes multiple additional interfaces (typically vlans off the primary interface) as well as multus networking. All networking can be IPv4 or IPv6. Pods may egress on the additional interfaces using routingViaHost and static routes added during cluster installation or via NMState.

      In order to limit the number of nodes, and thus IP Addresses needed, which are configured with the additional interfaces the design calls for the use of EgressIP. This ensures that traffic is routed to a limited number of "gateway nodes" where the additional interfaces are configured. 

      Within this configuration the customer has pods which must egress on multiple additional interfaces (not multus). In the current EgressIP feature there is support for only one EgressIP per pod which prevents its use with these pods.

      The request is to support:

      • Pods which egress on a "gateway" node through multiple additional interfaces
      • The selection of which additional interface is egressed is based on destination routing of the packet

      Note that support for multiple EgressIP associated with a single pod would meet the need under this RFE. The routing through a gateway node is the primary requirement. Association of a pod with an application specific IP is desired but not strictly required. This leaves open that a more limited implementation could be considered in which the traffic is forwarded to the "gateway" nodes for egress and destination based routing with SNAT to the nodes interface IP.

      3. Why does the customer need this? (List the business requirements here)

      Reduce the number of IP addresses allocated per network per cluster. In a production deployment of 7 clusters of 100 nodes each this will reduce the allocated IP addresses from 700 per additional network to ~70 (assuming 10 gateway nodes). 

      4. List any affected packages or components.

      ovn-k (EgressIP)

            mcurry@redhat.com Marc Curry
            rhn-support-imiller Ian Miller
            1 Vote for this issue
            13 Start watching this issue