Configuring switch failover for a Db2 pureScale environment on an InfiniBand network (Linux)

Switch failover capability is a high availability feature provided by the switch subnet manager (SM).

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 an InfiniBand (IB) 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 communication adapter ports on the primary and secondary cluster caching facilities (CFs).

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 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.

About this task

The procedure details steps for configuring multiple switches to support switch failover. Switch failover capability helps the resiliency, or fault tolerance, of a network.

To create a Db2 pureScale environment with multiple switches, you must have communication adapter ports on CF servers and you must configure switch failover on the switches.

Db2 pureScale environments with Linux systems and InfiniBand cluster interconnect networks require FabricIT EFM switch based fabric management software. For multiple communication adapter ports support on CF servers, the installation prerequisites list the minimum fabric manager image that must be installed on the switch (image-PPC_M405EX-EFM_1.1.2500.img). Depending on the fabric manager version required, a direct upgrade path to the minimum required version might not be supported. In cases where a direct upgrade path is not supported, you must upgrade the switch fabric manager software image by installing each subsequent version. For instructions on upgrading fabric manager software on a specific Mellanox switch, see the Mellanox website.

For single switch setup on InfiniBand networks, enabling the subnet manager (SM) on the switch is mandatory.


Restrictions

  1. Administrative access is required on the switches.

Procedure

  1. Upgrade the switch to the required fabric manager software version level listed in the Db2 pureScale Feature installation prerequisites.
    To obtain Fabric Manager software, see the Installation prerequisites for Db2 pureScale Feature (Intel Linux).
  2. Follow the steps for setting up the Mellanox subnet manager for a high availability domain (Mellanox SM HA), provided in the Mellanox switch user manual.

    With the Mellanox SM HA software, the system manager can enter and modify all IB subnet configurations for the subnet managers from a single location. Assign a virtual IP address (VIP) to the management port of the switch to manage the high availability domain. The system manager must configure all the switches in a Mellanox SM HA environment to join the same IB subnet, and assign the subnet a name. After joining the subnet, the subnet managers are synchronized and you must select one as the master subnet manager, and the others become standby subnet managers.

Example

Figure 1. Two switch SLES multiple cluster interconnects to CFs
The two CFs and four members connect on IP subnet 10.222.0.0 to a master and standby IB switch. A public Ethernet switch connects the two IB switches.

The public Ethernet switch is required to connect the two IB switches to configure the subnet.

What to do next

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