Switch configuration on a RoCE network (Linux)

Switch failover capability is a high availability feature provided by the configure Link Aggregate Control Protocol (LACP) on the switch.

Before you begin

The configuration procedure for switch failover detailed in this topic applies to both SuSE Linux® Enterprise Server (SLES) systems and Red Hat Enterprise Linux Server (RHEL) systems on a remote direct memory access (RDMA) over Converged Ethernet (RoCE) network.

This procedure does not apply to IBM® BladeCenter environments. To configure switch failover support for BladeCenter deployments, no actions are performed on the switch modules, but you must configure multiple cluster interconnects on the primary and secondary cluster caching facilities (CFs) and members.

For Db2® pureScale® environments that do not include blade servers, complete the following tasks before you begin:
  1. Set up your network topology according to one of the supported network topologies listed in Network topology configuration support for Db2 pureScale environments.
  2. Power on the switch and connect an RJ11 serial cable or Ethernet cable to the switch.

Administrative access is required on the switches.

About this task

A RoCE network switch must support:
  • Link Aggregate Control Protocol (LACP) for switch failover configuration
  • Global Pause flow control (IEEE 802.3x)
  • Optional. Support of local loopback IP address on the switch that can be pinged by IP addresses on the same IP subnet. For more details and restrictions on this, refer to the technote on Restrictions of automated adapter liveliness test.

This procedure configures multiple switches to support switch failover. Switch failover capability helps the resiliency or fault tolerance of a network. These steps are not applicable for a single switch setup.

To create a Db2 pureScale environment with multiple switches, you must have multiple cluster interconnects on CF and member servers and configure switch failover on the switches.


  1. Connect the two switches together.
    • Designate two or more ports on each switch to be used as inter-switch links (ISLs) and connect them physically.
    • Aggregate all ISLs using Link Aggregate Control Protocol (LACP).
    • All ISL ports on both switches must be setup as active.
  2. Disable the Converged Enhanced Ethernet (CEE) feature and Priority-Based Flow Control (PFC) (IEEE 802.1Qbb).
  3. Enable Global Pause flow control (IEEE 802.3x).
    • It involves configuration at the switch level only.
    • For a BNT switch with firmware level 6.5.2 and higher, Global Pause is enabled by setting both flow control send and receive to 'on' for all Db2 related ports including the ISL ports.
  4. Perform one of the following two Spanning Tree Protocol (STP) configurations.
    1. Disable STP
      • This reduces the overall configuration complexity and is suitable for dedicated private pureScale network where the pS switch is restricted to usage by pureScale hosts.
    2. Enable STP
      • This is recommended for shared pureScale network where non-pureScale traffic may go through the pureScale switches. Enabling STP can prevent accidental network loop creation. The actual commands vary with the switch brand and model. For BNT switch, the following settings are required:
        • Mark all non-ISL ports (including unused) as edge ports
        • Enable BPDU guard on all edge ports
        • Enable root guard on all edge ports
  5. Repeat the above on all switches in the cluster.
    Figure 1. Two CFs and four members connect to two switches.
    The two CFs and four members connect to two switches.

What to do next

Configure the network settings of hosts that you plan to include in the Db2 pureScale environment.